DETAILED CORRESPONDANCE
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of claims
This final office action on merits is in response to the communication received on 02/12/2026. Claims 40 and 41 are new. Amendments to claims 20, 23, and 30 are acknowledged and have been carefully considered. Claims 20-41 are pending and considered below.
Claim Rejections - 35 USC § 112
The amendment to claims 20 and 30 clarifies the relationship between the first threshold, second threshold, and cumulative metric, including how each value is determined and compared to generate a risk score. Accordingly, the indefiniteness issue previously identified under 35 U.S.C. § 112(b) has been overcome, and the rejection is withdrawn.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 20-39 are rejected under 35 U.S.C. 103 as being unpatentable over Shelton et al. (U.S. Patent Publication 2022/0237999 A1), referred to hereinafter as Shelton, in view of Burnett et al. (U.S. Patent Publication 2017/0136209 A1), referred to hereinafter as Burnett, Haber et al. (U.S. Patent Publication 2021/0142915 A1), referred to hereinafter as Haber, and Janssen et al. (U.S. Patent Publication 2021/0110902 A1), referred to hereinafter as Janssen.
Regarding claim 20, Shelton teaches a method for assessing risk of developing one or more post-operative complications for a target patient, the method comprising (Shelton [0006] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor that may be configured to perform the method. A biomarker for a patient may be received from a sensing system. A data collection may be received. The data collection may include pre-surgical data and surgical data for the patient. A risk assessment model may be determined using the data collection. The risk assessment model may be for a patient outcome associated with a surgery performed on the patient. A probability of a medical issue may be determined using the risk assessment model and the biomarker.” and Shelton [0059] “The sensing system may detect risk factors for postoperative complications including infection, cardiopulmonary complication, and/or bleeding episodes. Healthcare providers may use the detected risk factors for predicting or detecting post-operative or post-surgical complications, for example, to affect decisions and precautions taken during post-surgical care.”):
obtaining, by one or more processors in a server from at least one database, target patient historical data, historical data of a first plurality of patients ((Shelton [0330] “As shown in FIG. 9, the cloud system 20271 may include one or more central servers 20272 (may be same or similar to remote server 20067), surgical hub application servers 20276, data analytics modules 20277, and an input/output (“I/O”) interface 20278. The central servers 20272 of the cloud system 20271 may collectively administer the cloud computing system, which includes monitoring requests by client surgical hubs 20270 and managing the processing capacity of die cloud system 20271 for executing the requests. Each of the central servers 20272 may comprise one or more processors 20273 coupled to suitable memory devices 20274 which can include volatile memory such as random-access memory (RAM) and non-volatile memory such as magnetic storage devices. The memory devices 20274 may comprise machine executable instructions that when executed cause the processors 20273 to execute the data analytics modules 20277 for the cloud-based data analysis, real-time monitoring of measurement data received from the sensing systems 20268, operations, recommendations, and other operations as described herein. The processors 20273 can execute the data analytics modules 20277 independently or in conjunction with hub applications independently executed by the hubs 20270. The central servers 20272 also may comprise aggregated medical data databases 20275, which can reside in the memory 20274.”, and Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.”);
determining, by the one or more processors, a first set of risk scores associated with the target patient historical data using one or more historical data risk assessment models based, at least in part, on linking each of the first set of risk scores to a plurality of pre-operative risk factors comprising at least one of American Society of Anesthesiologists (AMA) scores, demographics, height, weight, surgical history, and vitals, wherein the one or more historical data risk assessment models are trained using the historical data of the first plurality of patients (Shelton [0472] “In an example, the risk assessment module 29072 may calculate outcome probabilities. The outcome probabilities may include a probability of a medical complication. The outcome probabilities may include a probability of a physiologic condition. The outcome probabilities may include a probability of a surgical outcome. The risk assessment module 29072 may calculate outcome probabilities based on data from the data collection 29069. The risk assessment module 29072 may calculate outcome probabilities based on data from the sensing systems, 29070 and 29076. The risk assessment module 29072 may calculate outcome probabilities based on data from the data interface 29071.”, Shelton [0473] “In an example, the risk assessment module 29072 may calculate outcome probabilities based on a risk model. The risk model may include a data trend, table, artificial intelligence model, and the like. The risk model may be obtained. The risk model may be obtained based on at least one of patient data, biomarker data, EMR, and the like. The risk model may provide a situation awareness.”, Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.”, and Shelton [0006] “The data collection may include pre-surgical data and surgical data for the patient. A risk assessment model may be determined using the data collection. The risk assessment model may be for a patient outcome associated with a surgery performed on the patient.”, and Shelton [0368] “Model training may be another aspect of the machine learning lifecycle. The model training process as described herein may be dependent on the machine learning algorithm used. A model may be deemed suitably trained after it has been trained, cross validated, and tested. Accordingly, the dataset from the data preparation stage (e.g., an input dataset) may be divided into a training dataset (e.g., 60% of the input dataset), a validation dataset (e.g., 20% of the input dataset), and a test dataset (e.g., 20% of the input dataset). After the model has been trained on the training dataset, the model may be run against the validation dataset to reduce overfitting. If accuracy of the model were to decrease when run against the validation dataset when accuracy of the model has been increasing, this may indicate a problem of overfitting. The test dataset may be used to test the accuracy of the final model to determine whether it is ready for deployment or more training may be required.”);
determining, by the one or more processors, a second set of risk scores associated with the real-time data continuously using one or more real-time risk assessment models (Shelton [0006] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor that may be configured to perform the method. A biomarker for a patient may be received from a sensing system. A data collection may be received. The data collection may include pre-surgical data and surgical data for the patient. A risk assessment model may be determined using the data collection. The risk assessment model may be for a patient outcome associated with a surgery performed on the patient. A probability of a medical issue may be determined using the risk assessment model and the biomarker.” or Shelton [0106] “Based on the measured edema data, the edema sensing system may detect edema-related biomarkers, complications, and/or contextual information, including inflammation, rate of change in inflammation, poor healing, infection, leak, colorectal anastomotic leak, and/or water build-up.” Shelton [0107] “For example, the edema sensing system may detect a risk of colorectal anastomotic leak based on fluid build-up. Based on the detected edema physiological conditions, the edema sensing system may generate a score for healing quality. For example, the edema sensing system may generate the healing quality score by comparing edema information to a certain threshold lower leg circumference. Based on the detected edema information, the edema sensing system may generate edema tool parameters including responsiveness to stapler compression. The edema sensing system may provide context for measured edema data by using measurements from the accelerometer, gyroscope, and/or magnetometer. For example, the edema sensing system may detect whether the user is sitting, standing, or lying down.”, and Shelton [0254] “Applying cloud computer data processing techniques on the measurement data collected by the sensing systems 20069, the surgical data network can provide improved surgical outcomes, improved recovery outcomes, reduced costs, and improved patient satisfaction. At least some of the sensing systems 20069 may be employed to assess physiological conditions of a surgeon operating on a patient or a patient being prepared for a surgical procedure or a patient recovering after a surgical procedure. The cloud-based computing system 20064 may be used to monitor biomarkers associated with a surgeon or a patient in real-time and to generate surgical plans based at least on measurement data gathered prior to a surgical procedure, provide control signals to the surgical instruments during a surgical procedure, notify a patient of a complication during post-surgical period.”),
wherein the one or more real-time risk assessment models are trained by ((Shelton [0450] “In an example, the previous patient outcomes and/or risk models may be trained. The trained risk models may provide more accurate data for analysis. Machine learning may be provided to train the previous patient outcomes and/or risk models. The machine learning may be supervised and/or unsupervised, for example. The machine learning may include updating the previous patient outcomes data and risk models with recent analyses. The machine learning's training process may update the risk model for predicting an outcome on new data set. For example, previous patient outcomes data and risk models may be updated with new analyses and determinations relating to risk of septic heart rate. The machine learning may use the determined risk of complication to train the risk model, for example. After an analysis, the previous patient outcomes and/or risk models may be updated to provide more accurate data for future analysis.” Shelton [0006] “The data collection may include pre-surgical data and surgical data for the patient. A risk assessment model may be determined using the data collection. The risk assessment model may be for a patient outcome associated with a surgery performed on the patient.”, Shelton [0368] “Model training may be another aspect of the machine learning lifecycle. The model training process as described herein may be dependent on the machine learning algorithm used. A model may be deemed suitably trained after it has been trained, cross validated, and tested. Accordingly, the dataset from the data preparation stage (e.g., an input dataset) may be divided into a training dataset (e.g., 60% of the input dataset), a validation dataset (e.g., 20% of the input dataset), and a test dataset (e.g., 20% of the input dataset). After the model has been trained on the training dataset, the model may be run against the validation dataset to reduce overfitting. If accuracy of the model were to decrease when run against the validation dataset when accuracy of the model has been increasing, this may indicate a problem of overfitting. The test dataset may be used to test the accuracy of the final model to determine whether it is ready for deployment or more training may be required.”); and
generating, by the one or more processors, a risk assessment based on the first set of risk scores and the second set of risk scores of the target patient developing the one or more post-operative complications to inform a post-operative treatment plan (Shelton [0493] “Risk assessment modeling may be provided. Risk assessment modeling may indicate a probability of issue development from a surgery. The probability of issue development from a surgery may be indicated based on measured biomarkers of a patient before and after surgery. The risk assessment modeling may include a wearable that incorporates pre-operative and/or post-operative data in a risk model with post-operative data. The risk assessment modeling may include coupling biomarkers for discernment between a healthy and concerning output of a biomarker”, Shelton [0045] “The computing system may generate a treatment plan, including a pain management strategy, based on the sleep biomarkers. The surgical hub may detect potential risk factors or conditions, including systemic inflammation and/or reduced immune function, based on the sleep biomarkers.”, and Shelton [0118] “Based on the detected circulating tumor cells data, the circulating tumor cell sensing system may calculate metastatic risk. The presence of circulating cancerous cells may indicate metastatic risk. Circulating cancerous cells per milliliter of blood exceeding a threshold amount may indicate a metastatic risk. Cancerous cells may circulate the bloodstream when tumors metastasize. Based on the calculated metastatic risk, the circulating tumor cell sensing system may generate a surgical risk score. Based on the generated surgical risk score, the circulating tumor cell sensing system may indicate surgery viability and/or suggested surgical precautions.”).
Shelton fails to explicitly teach time marked data of a second plurality of patients; measuring, by an inline sensor device, real-time data at regular intervals based on biomarker signals associated with bodily fluids from the target patient, wherein the inline sensor device comprises an input port in fluid communication with the bodily fluids from the target patient via a catheter, an output port in fluid communication with a reservoir, a fluid channel configured for passing the bodily fluids, one or more sensors embedded in the fluid channel configured for measuring the biomarker signals associated with the bodily fluids flowing across the one or more sensors through the fluid channel from the input port to the output port, an a connection mechanism configured for sending the real-time data to the server; using the time-marked data of the second plurality of patients; determining a first threshold for each post-operative time interval from the time-marked data, the first threshold distinguishing patients associated with one or more post-operative complications from patients not associated with the one or more post-operative complications; determining a second threshold using a temporal filter applied to a percentage of the time-marked data below the first threshold; and determining a cumulative metric by applying the temporal filter to a percentage of the real-time data below the first threshold: and comparing the cumulative metric and the second threshold, wherein the cumulative metric above the second threshold indicates a high risk score, and the cumulative metric below the second threshold indicates a low risk score.
Burnett teaches time marked data of a second plurality of patients (Burnett [0013] “The disclosed technology also relates to a Foley type catheter for sensing physiologic data from the bladder and/or urinary tract of a patient, the physiologic data particularly including those gathered by high fidelity pressure sensing and transduction into signals suitable for processing. In some embodiments, the pressure-sensing Foley type catheter may further be enabled to sense temperature and analytes of clinical significance. Examples of physiological parameters that the sensing Foley catheter system may measure (time specific measurements and trends of values over time) include: urine output, respiration rate, heart rate, heart rate variability, stroke volume, stroke volume variability, intra-abdominal pressure (IAP), tissue oxygenation, tissue gas content, pulse transit time, pulmonary blood volume variability, temperature, blood content and other patient parameters”);
measuring, by an inline sensor device, real-time data at regular intervals based on biomarker signals associated with bodily fluids from the target patient (Burnett “A pressure measuring balloon on a catheter, such as that disclosed in international patent application number PCT/US14/44565, titled Sensing Foley Catheter (which is herein incorporated by reference in its entirety) is an example of a device which measures patient parameters. Additional embodiments are disclosed herein. A sensing Foley catheter system, may include a pressure measuring balloon and/or other sensors, as well as the ability to measure urine output and content to determine patient parameters such as urine output rate, IAP, respiratory rate, heart rate, stroke volume, tissue oxygenation, urine composition, temperature and other patient parameters.” and Burnett [0182] “Other wavelengths and other technologies may also be used to detect various substances in urine or any collected/drained bodily fluid. UV/light absorption may also be used to detect turbidity. A dye or drug or reactive substance may also be introduced into the system, or be coated on the inside of the system, cassette, etc, to react with a substance in the urine to aid in analysis. Any type of sensor may be used to sense any substance or quality of the collected urine in either an intermittent or continuous basis, real-time basis. For example, sensor(s) to detect Magnesium in the urine may be used to diagnose preeclampsia or eclampsia. Lactate sensors may be used to test for lactate (or lactate dehydrogenase) in the urine. The identification of lactate in urine may be an early indicator of sepsis. Lactate sensors may include enzymatic lactate sensors. For example, lactate sensors as disclosed in Weber (Weber J., Kumar A., Kumar A., Bhansali S. Novel lactate and pH biosensor for skin and sweat analysis based on single walled carbon nanotubes. Sens. Actuators, B. Chem. 2006; 117:308-313), and/or Mo (Mo, J W, Smart, W. Lactate biosensors for continuous monitoring. Front Biosci. 2004 Sep. 1; 9:3384-91), both of which are incorporated herein by reference in their entirety, may be used.”),
wherein the inline sensor device comprises an input port in fluid communication with the bodily fluids from the target patient via a catheter (Burnett [0137] “Sensing Foley catheter 1000 is similar to the sensing Foley catheter shown in FIG. 1. The sensing Foley catheter is shown in use in bladder 1014. Note that several of the ports at the proximal end of the catheter shown in FIG. 1 are combined in the embodiment shown in FIG. 10A. Urine drainage tube 1001 is also shown here. The urine drainage tube may be combined with the sensing Foley catheter or may be a separate component. Urine drainage tube 1001 and/or sensing Foley catheter may also include vent barb 1016, or the vent barb may be a separate component. Airlock clearing mechanism and fluid collection & analysis system 1002 is also shown here, and is in fluid communication with urine drainage tube 1001 which is in fluid communication with sensing Foley catheter 1000. Airlock clearing mechanism and fluid collection & analysis system includes base/controller 1018, fluid collection bag 1020 and reservoir or cassette 1022. The combination of the sensing Foley catheter 1000, the urine drainage tube 1001, and the airlock clearing mechanism and fluid collection & analysis system 1002 are also referred to herein as the sensing Foley catheter system. The sensing Foley catheter, urine drainage line, and reservoir/cassette may be disposable and may be sold as a unit. This disposable assembly is shown in FIG. 10C, which includes sensing Foley catheter 1000, urine drainage tube 1001 (including vent barb) and reservoir/cassette 1022.”),
an output port in fluid communication with a reservoir (Burnett [0148] “Urine/fluid drainage bag 1020 includes one way valves 1136 connected to overflow tubing 1138 and outflow tubing 1140 to prevent urine/fluid from exiting the drainage bag once collected. These valves also prevent air from entering the collection vessel 1022 when pump 1134 is pulling vacuum so that the vacuum acts on the drainage tubing and not the bag. In a preferred embodiment, a single valve is used for both the overflow and outflow tubings. Mounting hooks/holes 1102 allow drainage bag 1020 to be removably attached to controller 1018. Vent 1142, which may be a hydrophobic or other vent, allows air or gas to exit the drainage bag, but does not allow fluid to exit the bag. This prevents excessive air, and potentially pressure, buildup in the bag, and thus allows for efficient filling of the drainage bag. Graduated markings 1144 show a somewhat crude measurement of the fluid volume in the bag as it is collected. Outflow valve 1146 may be used to empty the bag of fluid/urine. Preferably, the valve is operable easily by one person. Collection bag hooks 1102 when designed as strain measurement elements may also force an alarm to sound if the bag is reaching full capacity and needs to be emptied. An alarm may also sound if there is unnecessarily excessive force on the bag, for example if the bag is being pulled or is caught on an obstacle as a patient is being moved.”),
a fluid channel configured for passing the bodily fluids, one or more sensors embedded in the fluid channel configured for measuring the biomarker signals associated with the bodily fluids flowing across the one or more sensors through the fluid channel from the input port to the output port (Burnett [0137] “Sensing Foley catheter 1000 is similar to the sensing Foley catheter shown in FIG. 1. The sensing Foley catheter is shown in use in bladder 1014. Note that several of the ports at the proximal end of the catheter shown in FIG. 1 are combined in the embodiment shown in FIG. 10A. Urine drainage tube 1001 is also shown here. The urine drainage tube may be combined with the sensing Foley catheter or may be a separate component. Urine drainage tube 1001 and/or sensing Foley catheter may also include vent barb 1016, or the vent barb may be a separate component. Airlock clearing mechanism and fluid collection & analysis system 1002 is also shown here, and is in fluid communication with urine drainage tube 1001 which is in fluid communication with sensing Foley catheter 1000. Airlock clearing mechanism and fluid collection & analysis system includes base/controller 1018, fluid collection bag 1020 and reservoir or cassette 1022. The combination of the sensing Foley catheter 1000, the urine drainage tube 1001, and the airlock clearing mechanism and fluid collection & analysis system 1002 are also referred to herein as the sensing Foley catheter system. The sensing Foley catheter, urine drainage line, and reservoir/cassette may be disposable and may be sold as a unit. This disposable assembly is shown in FIG. 10C, which includes sensing Foley catheter 1000, urine drainage tube 1001 (including vent barb) and reservoir/cassette 1022.” and Burnett [0020] “A pressure measuring balloon on a catheter, such as that disclosed in international patent application number PCT/US14/44565, titled Sensing Foley Catheter (which is herein incorporated by reference in its entirety) is an example of a device which measures patient parameters. Additional embodiments are disclosed herein. A sensing Foley catheter system, may include a pressure measuring balloon and/or other sensors, as well as the ability to measure urine output and content to determine patient parameters such as urine output rate, IAP, respiratory rate, heart rate, stroke volume, tissue oxygenation, urine composition, temperature and other patient parameters.”), and
a connection mechanism configured for sending the real-time data to the server (Burnett [0145] “The patient data may be transferred wirelessly or by wired connection to another storage device, such as a server on the internet or an intranet. Patient data may be anonymized. Patient data, such as the patient ID, may be stored in an RFID adapter so that data specific to a particular patient is recognized by the controller and associated with the disposable component used by that patient.”, and Burnett [0212] “FIG. 33 is a detailed diagram of the loop controller. Loop controller 2928 can receive one or more patient parameter inputs from a sensing Foley catheter or other device. These inputs include, but are not limited to, urine output volume and rate, pressure profile from the bladder, and sensor info from a sensing Foley catheter or other device. Pressure profile info from the bladder can be further analyzed to determine IAP, respiratory rate, heart rate, stroke volume, sepsis index, AKI index and other patient parameters. This analysis may be performed in loop controller 2928 or in a separate controller which is connected to loop controller either by a wired or wireless connection. The connection may be via an internet, intranet, WAN, LAN or other network, or it may be local via Bluetooth, Wi-Fi, etc.”);
using the time-marked data of the second plurality of patients (Burnett [0013] “The disclosed technology also relates to a Foley type catheter for sensing physiologic data from the bladder and/or urinary tract of a patient, the physiologic data particularly including those gathered by high fidelity pressure sensing and transduction into signals suitable for processing. In some embodiments, the pressure-sensing Foley type catheter may further be enabled to sense temperature and analytes of clinical significance. Examples of physiological parameters that the sensing Foley catheter system may measure (time specific measurements and trends of values over time) include: urine output, respiration rate, heart rate, heart rate variability, stroke volume, stroke volume variability, intra-abdominal pressure (IAP), tissue oxygenation, tissue gas content, pulse transit time, pulmonary blood volume variability, temperature, blood content and other patient parameters”):
Haber teaches determining a first threshold for each post-operative interval, the first threshold distinguishing patients associated with one or more post-operative complications from patients not associated with the one or more post-operative complications (Haber [0086] “However, in an alternative illustrative implementation, an alerting and attribution algorithm development process 294 receives as inputs, the baseline risk variables 268, dynamic risk variables, 270, the baseline outcome likelihood model 288, dynamic outcome likelihood model 290 and the second training data set 284 (e.g., outcome and non-outcome data) to output one or more alerting and attribution algorithms 296. In illustrative implementations, the alerts generated by the alerting and attribution algorithm development process 294 are generated based upon a comparison of dynamic and baseline risk in view of available patient data, and are characterized in terms of sensitivity and specificity, which are embodied in a fixed-threshold alert value. In an example implementation, the user can select an alerting threshold that balances the competing goals of minimizing false negatives and false positives. Receiver-operator characteristic curves are used to characterize true positive/false positive behavior for all possible thresholds and to aid in threshold selection. For instance, the formula variables (values), including predicted outcome data Ŷ0 . . . Ŷn and actual outcomes can be used to set the threshold.” and Haber [0088] “As noted in greater detail herein, an alert provides information to a clinical staff in response to the system detecting that a potentially preventable adverse outcome is likely to occur in the future. Thus, “time to risk” is a factor that may be accounted for in the model development process. For instance, the outcome may be predicted in some finite time, e.g., 48 hours from a particular event, at some time during the patient hospital stay, etc. As such, the processing is more sophisticated than mere aggregation.”).
Janssen teaches a first threshold for each time interval from the time-marked data (Janssen [0033] FIG. 5 is another representative image showing the monitored data obtained from a patient over the same four-day monitoring and analysis period. The trace 46, which is an ECG artifact-based example, represents an exemplary time series smoothing approach with a twenty-minute moving average of the total alert frequency/count while trace 47 is a sixty-minute moving average of the total time in alert shown over the four-day monitoring and analysis period. In the monitor alert interpretation display 48 of FIG. 5, a determined artifact-specific enhanced threshold 50a is shown for the twenty-minute moving average, which is set at sixty-five alerts per twenty minute threshold. The enhanced threshold 50b is shown for the sixty-minute moving average of the total time in alert and is set at 2.5 minutes of artifact alert per the sixty minute threshold. In addition to the traces 46 and 47, lead failures 51 are shown in the upper portion 49 of the display. In the time periods shown by reference numeral 52 and reference numeral 61, a significant number of artifacts (i.e., over sixty-five artifact alerts per twenty minutes in the illustrated example) are detected and may be indicated to the monitoring technician each time they occur.”);
determining a second threshold using a temporal filter applied to a percentage of the time-marked data below the first threshold (Janssen [0033] FIG. 5 is another representative image showing the monitored data obtained from a patient over the same four-day monitoring and analysis period. The trace 46, which is an ECG artifact-based example, represents an exemplary time series smoothing approach with a twenty-minute moving average of the total alert frequency/count while trace 47 is a sixty-minute moving average of the total time in alert shown over the four-day monitoring and analysis period. In the monitor alert interpretation display 48 of FIG. 5, a determined artifact-specific enhanced threshold 50a is shown for the twenty-minute moving average, which is set at sixty-five alerts per twenty minute threshold. The enhanced threshold 50b is shown for the sixty-minute moving average of the total time in alert and is set at 2.5 minutes of artifact alert per the sixty minute threshold. In addition to the traces 46 and 47, lead failures 51 are shown in the upper portion 49 of the display. In the time periods shown by reference numeral 52 and reference numeral 61, a significant number of artifacts (i.e., over sixty-five artifact alerts per twenty minutes in the illustrated example) are detected and may be indicated to the monitoring technician each time they occur. The present disclosure addresses this issue by providing a contextual enhanced alert or “meta-alert” and supporting visual aids identifying these sustained alert cases for each parameter or combinations of parameters. In the embodiment shown in FIG. 5, the enhanced threshold 50a is set for the moving average of the number of artifacts that occur over a selected period of time. As an illustrative example, the enhanced threshold 50a could be set at an intensity of sixty-five occurrences over a twenty minute period. It should be understood that the enhanced threshold 50a is set separately for each type of alert being monitored and that the intensity could be changed/modified by each facility. When the tracked moving average exceeds the enhanced threshold 50a, an enhanced alert is generated. Such enhanced alert will allow the monitoring technician who is managing multiple patients to recognize the repeated generation of individual artifact alerts, understand the longitudinal context of the frequency of the alerts, individual alert duration, the accumulated duration of the alerts, and the potential repeated and sustained nature of the alerts. Without such an enhanced alert, it is very difficult for the monitoring technician to grasp the frequency and duration of the artifact alerts.”, and Janssen [0050] “In step 78, the data analysis module 64 performs ongoing data analysis on the monitored data obtained from each of the individual patients. As described previously, the data analysis module 64 is able to carry out a wide variety of analysis on the monitored data, which can include alarm/alert frequency obtained from the patient. In an exemplary case, the data analysis module can determine in step 78 periods of no telemetry, when individual leads have failed, when arrhythmia detection has been suspended, when the system is the relearning process and periods when relatively significant artifacts are present. Based upon the analysis of the data in step 78, the method moves to step 80 in which the data analysis module compares the information to rulesets obtained from the ruleset module 66. As indicated previously, in step 80, the data analysis module 64 compares the number, frequency and/or combined duration of threshold alerts that occur to a series of enhanced alert thresholds that are obtained from the ruleset module 66. An enhanced alert threshold is a threshold that indicates when a parameter's data series is analyzed for frequency, time/duration, etc. in alert via a smoothing algorithm or machine learning and exceeds a determined value over time. Alternatively, the aggregated amount of time that an alert is being generated over the total number of alerts can be monitored to provide another basis for generating an alert to the clinician.”):
determining a cumulative metric by applying the temporal filter to a percentage of the real-time data below the first threshold: and comparing the cumulative metric and the second threshold, wherein the cumulative metric above the second threshold indicates a high risk score, and the cumulative metric below the second threshold indicates a low risk score (Janssen [0033] FIG. 5 is another representative image showing the monitored data obtained from a patient over the same four-day monitoring and analysis period. The trace 46, which is an ECG artifact-based example, represents an exemplary time series smoothing approach with a twenty-minute moving average of the total alert frequency/count while trace 47 is a sixty-minute moving average of the total time in alert shown over the four-day monitoring and analysis period. In the monitor alert interpretation display 48 of FIG. 5, a determined artifact-specific enhanced threshold 50a is shown for the twenty-minute moving average, which is set at sixty-five alerts per twenty minute threshold. The enhanced threshold 50b is shown for the sixty-minute moving average of the total time in alert and is set at 2.5 minutes of artifact alert per the sixty minute threshold. In addition to the traces 46 and 47, lead failures 51 are shown in the upper portion 49 of the display. In the time periods shown by reference numeral 52 and reference numeral 61, a significant number of artifacts (i.e., over sixty-five artifact alerts per twenty minutes in the illustrated example) are detected and may be indicated to the monitoring technician each time they occur. The present disclosure addresses this issue by providing a contextual enhanced alert or “meta-alert” and supporting visual aids identifying these sustained alert cases for each parameter or combinations of parameters. In the embodiment shown in FIG. 5, the enhanced threshold 50a is set for the moving average of the number of artifacts that occur over a selected period of time. As an illustrative example, the enhanced threshold 50a could be set at an intensity of sixty-five occurrences over a twenty minute period. It should be understood that the enhanced threshold 50a is set separately for each type of alert being monitored and that the intensity could be changed/modified by each facility. When the tracked moving average exceeds the enhanced threshold 50a, an enhanced alert is generated. Such enhanced alert will allow the monitoring technician who is managing multiple patients to recognize the repeated generation of individual artifact alerts, understand the longitudinal context of the frequency of the alerts, individual alert duration, the accumulated duration of the alerts, and the potential repeated and sustained nature of the alerts. Without such an enhanced alert, it is very difficult for the monitoring technician to grasp the frequency and duration of the artifact alerts.”, and Janssen [0050] “In step 78, the data analysis module 64 performs ongoing data analysis on the monitored data obtained from each of the individual patients. As described previously, the data analysis module 64 is able to carry out a wide variety of analysis on the monitored data, which can include alarm/alert frequency obtained from the patient. In an exemplary case, the data analysis module can determine in step 78 periods of no telemetry, when individual leads have failed, when arrhythmia detection has been suspended, when the system is the relearning process and periods when relatively significant artifacts are present. Based upon the analysis of the data in step 78, the method moves to step 80 in which the data analysis module compares the information to rulesets obtained from the ruleset module 66. As indicated previously, in step 80, the data analysis module 64 compares the number, frequency and/or combined duration of threshold alerts that occur to a series of enhanced alert thresholds that are obtained from the ruleset module 66. An enhanced alert threshold is a threshold that indicates when a parameter's data series is analyzed for frequency, time/duration, etc. in alert via a smoothing algorithm or machine learning and exceeds a determined value over time. Alternatively, the aggregated amount of time that an alert is being generated over the total number of alerts can be monitored to provide another basis for generating an alert to the clinician.”).
It would have been obvious to a PHOSITA before the effective filing date of the invention to combine the teachings of Shelton, Burnett, Haber, and Janssen to arrive at the claimed postoperative complication risk assessment method because the references address aspects of physiological monitoring, predictive risk modeling, and threshold alert generation in clinical environments. Shelton teaches cloud surgical monitoring systems that collect historical and real-time biomarker data, apply machine learning risk assessment models trained on patient datasets, generate complication probabilities and risk scores, and use the resulting assessments to guide postoperative treatment and clinical decision-making. Burnett teaches inline Foley catheter sensing systems capable of continuously measuring physiological and biomarker data from bodily fluids in real time through fluid channels and transmitting the resulting data to remote servers for analysis. Haber teaches generating thresholds from outcome and non-outcome patient datasets using predictive modeling and ROC threshold selection for identifying adverse outcomes over clinically relevant time periods. Janssen teaches applying temporal filtering, moving averages, and enhanced thresholding to continuously monitored physiological time-series data in order to evaluate sustained conditions and generate clinically meaningful alerts over time.
A PHOSITA would have been motivated to integrate Burnett’s inline bodily-fluid sensing architecture into Shelton’s cloud postoperative risk assessment framework because Shelton focuses on incorporating biomarker sensing systems and real-time physiological monitoring into predictive surgical outcome models, while Burnett provides a directly compatible mechanism for continuously acquiring biomarker measurements from catheterized patients and transmitting those measurements for server-side analysis. The combination would have predictably improved the quantity of physiological biomarker inputs available to Shelton’s machine learning risk assessment models, which would improving the accuracy and responsiveness of postoperative complication detection and risk prediction. Both references are directed to improving postoperative patient monitoring using continuously acquired physiological measurements and analysis, and their combination represents the predictable use of known sensing techniques within an existing risk assessment architecture.
Further, a PHOSITA would have been motivated to incorporate Haber’s threshold-generation techniques and Janssen’s temporal filtering and cumulative alert processing into the combined Shelton and Burnett system to improve the reliability of continuously generated postoperative risk assessments. Haber teaches generating thresholds using outcome and non-outcome datasets and accounting for “time to risk” in predictive modeling, which suggest threshold determination using time-associated patient data for distinguishing complication states. Janssen further teaches applying smoothing algorithms, moving averages and enhanced thresholds to physiological monitoring data over defined monitoring intervals in order to identify sustained physiological conditions while reducing transient noise and false alerts. A PHOSITA would have recognized that applying Janssen’s temporal filtering and cumulative thresholding techniques to the continuously acquired physiological time-series data of Burnett within Shelton’s machine learning risk assessment framework would have predictably improved alert stability and enabled more reliable continuous postoperative complication monitoring over time.
Additionally, a PHOSITA would have understood that the use of separate historical-data and real-time-data risk assessment models, trained using different datasets and updated with continuously acquired physiological measurements, which reflect the application of known machine learning training and monitoring techniques to improve predictive accuracy in a clinical monitoring environment. Shelton teaches model training, validation, updating of risk models using prior patient outcome data, and continuous monitoring of biomarkers before and after surgery. Haber and Janssen further reinforce the use of temporally analyzed patient monitoring data and threshold optimization techniques for improving predictive assessments. Accordingly, combining the references would have yielded no more than the predictable use of prior art elements according to their established functions to improve postoperative risk prediction and clinical monitoring systems.
Regarding claim 21, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach wherein the target patient historical data comprises demographics and medical history associated with the target patient (Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to use patient demographics and medical history as part of the target patient historical data, as taught by Shelton, because Shelton discloses that EMRs include patient demographics, medical history, past procedures, medications, and other data relevant to surgical outcome prediction. Shelton provides a clear teaching, suggestion, and motivation (TSM) to incorporate these forms of historical data into risk modeling. Accordingly, one of ordinary skill would have been motivated to include demographics and medical history when obtaining historical data for the claimed risk assessment method, since this data are conventionally used to improve surgical risk prediction and have a predictable effect on model performance.
Regarding claim 22, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach wherein the one or more historical data risk assessment models are trained by (Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.” and Shelton [0006] “The data collection may include pre-surgical data and surgical data for the patient. A risk assessment model may be determined using the data collection. The risk assessment model may be for a patient outcome associated with a surgery performed on the patient.”):
obtaining, by the one or more processors, the historical data corresponding to the first plurality of patients, wherein the historical data provides an indication of whether each patient of the first plurality of patients encountered the one or more post- operative complications (Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.” and Shelton [0005] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor. The processor may be configured to perform the method. A biomarker may be received for a patient from a sensing system. A data collection that includes pre-surgical data may be received for a patient. A probability of a patient outcome due to a surgery performed on the patient may be determined using the biomarker and the data collection. A notification may be sent to a user. The notification may indicate that the probability of the surgical complication may exceed a threshold.”);
generating pre-processed historical data by (Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.” and Shelton [0366] “Data preparation may be performed for machine learning as another stage of the machine learning lifecycle. Data preparation may include data preprocessing steps such as data formatting, data cleaning, and data sampling.”):
mapping, via a mapping engine of the one or more processors, the historical data for each patient of the first plurality of patients to numerical values (Shelton [0366] “Data preparation may be performed for machine learning as another stage of the machine learning lifecycle. Data preparation may include data preprocessing steps such as data formatting, data cleaning, and data sampling. For example, the collected data may not be in a data format suitable for training a model. In an example, a patient's integrated data record of pre-surgical EMR record data and biomarker measurement data, surgical data, and post-surgical data may be in a rational database. Such data record may be converted to a flat file format for model training. In an example, the patient's pre-surgical EMR data may include medical data in text format, such as the patient's diagnoses of emphysema, pre-operative treatment (e.g., chemotherapy, radiation, blood thinner). Such data may be mapped to numeric values for model training. For example, the patient's integrated data record may include personal identifier information or other information that may identifier a patient such as an age, an employer, a body mass index (BMI), demographic information, and the like. Such identifying data may be removed before model training. For example, identifying data may be removed for privacy reasons. As another example, data may be removed because there may be more data available than may be used for model training. In such case, a subset of the available data may be randomly sampled and selected for model training and the remainder may be discarded.”); and
standardizing, via a standardization engine of the one or more processors, the numerical values to have a mean of 0 and a standard deviation of 1 (Shelton [0367] “Data preparation may include data transforming procedures (e.g., after preprocessing), such as scaling and aggregation. For example, the preprocessed data may include data values in a mixture of scales. These values may be scaled up or down, for example, to be between 0 and 1 for model training. For example, the preprocessed data may include data values that carry more meaning when aggregated. In an example, there may be multiple prior colorectal procedures a patient has had. The total count of prior colorectal procedures may be more meaningful for training a model to predict surgical complications due to adhesions. In such case, the records of prior colorectal procedures may be aggregated into a total count for model training purposes.”):
inputting, by the one or more processors, the pre-processed historical data into one or more historical data risk assessment models (Shelton [0368] “Model training may be another aspect of the machine learning lifecycle. The model training process as described herein may be dependent on the machine learning algorithm used. A model may be deemed suitably trained after it has been trained, cross validated, and tested. Accordingly, the dataset from the data preparation stage (e.g., an input dataset) may be divided into a training dataset (e.g., 60% of the input dataset), a validation dataset (e.g., 20% of the input dataset), and a test dataset (e.g., 20% of the input dataset). After the model has been trained on the training dataset, the model may be run against the validation dataset to reduce overfitting. If accuracy of the model were to decrease when run against the validation dataset when accuracy of the model has been increasing, this may indicate a problem of overfitting. The test dataset may be used to test the accuracy of the final model to determine whether it is ready for deployment or more training may be required.”); and
performing, by the one or more processors, one or more regression analyses on the pre-processed historical data to determine a relationship between the pre-processed historical data and the one or more risk scores (Shelton [0364] “The output of machine learning's training process may be a model for predicting outcome(s) on a new dataset. For example, a linear regression learning algorithm may be a cost function that may minimize the prediction errors of a linear prediction function during the training process by adjusting the coefficients and constants of the linear prediction function. When a minimal may be reached, the linear prediction function with adjusted coefficients may be deemed trained and constitute the model the training process has produced.” And Shelton [0360] “Machine learning may be supervised (e.g., supervised learning). A supervised learning algorithm may create a mathematical model from training a dataset (e.g., training data). The training data may consist of a set of training examples. A training example may include one or more inputs and one or more labeled outputs. The labeled output(s) may serve as supervisory feedback. In a mathematical model, a training example may be represented by an array or vector, sometimes called a feature vector. The training data may be represented by row(s) of feature vectors, constituting a matrix. Through iterative optimization of an objective function (e.g., cost function), a supervised learning algorithm may learn a function (e.g., a prediction function) that may be used to predict the output associated with one or more new inputs. A suitably trained prediction function may determine the output for one or more inputs that may not have been a part of the training data. Example algorithms may include linear regression, logistic regression, and neutral network. Example problems solvable by supervised learning algorithms may include classification, regression problems, and the like.”, Shelton [0118] “Based on the detected circulating tumor cells data, the circulating tumor cell sensing system may calculate metastatic risk. The presence of circulating cancerous cells may indicate metastatic risk. Circulating cancerous cells per milliliter of blood exceeding a threshold amount may indicate a metastatic risk. Cancerous cells may circulate the bloodstream when tumors metastasize. Based on the calculated metastatic risk, the circulating tumor cell sensing system may generate a surgical risk score. Based on the generated surgical risk score, the circulating tumor cell sensing system may indicate surgery viability and/or suggested surgical precautions.”, Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.” and Shelton [0005] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor. The processor may be configured to perform the method. A biomarker may be received for a patient from a sensing system. A data collection that includes pre-surgical data may be received for a patient. A probability of a patient outcome due to a surgery performed on the patient may be determined using the biomarker and the data collection. A notification may be sent to a user. The notification may indicate that the probability of the surgical complication may exceed a threshold.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to train the historical data risk assessment models using known machine learning preprocessing and training steps, as Shelton teaches obtaining EMR historical data, mapping data to numerical values for model training, cleaning and formatting data, scaling and standardizing values, and performing regression based supervised learning. Shelton provides express motivation to prepare EMR data in this manner to generate predictive risk models for surgical complications. Given this teaching, one of ordinary skill would have been motivated to follow the conventional steps disclosed by Shelton, collecting historical labeled records, numerically encoding categorical data, standardizing values to common scales, and training regression-based models, to yield a predictive historical risk model, because these are routine steps in supervised machine learning pipelines designed to produce accurate clinical outcome models.
Regarding claim 23, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach wherein the one or more real-time risk assessment models are trained by (Shelton [0354] “In an example, the surgical hub 20326 may be communicably coupled to one or more surgeon and/or patient sensing systems 20336. The sensing systems 20336 may be used to measure and/or monitor, in real-time, various biomarkers associated with a surgeon performing a surgical procedure or a patient on whom a surgical procedure is being performed. A list of the patient/surgeon biomarkers measured by the sensing systems 20336 is provided herein. In an example, the surgical hub 20326 may be communicably coupled to an environmental sensing system 20334. The environmental sensing systems 20334 may be used to measure and/or monitor, in real-time, environmental attributes, for example, temperature/humidity in the surgical theater, surgeon movements, ambient noise in the surgical theater caused by the surgeon's and/or the patient's breathing pattern, etc.” and Shelton [0373] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor. The processor may be configured to perform the method. A biomarker may be received for a patient from a sensing system. A data collection that includes pre-surgical data may be received for a patient. A probability of a patient outcome due to a surgery performed on the patient may be determined using the biomarker and the data collection. A notification may be sent to a user. The notification may indicate that the probability of the surgical complication may exceed a threshold.”):
obtaining, by the one or more processors, the time-marked data corresponding to the second plurality of patients, wherein the time-marked data provides an indication of whether each patient of the second plurality of patients encountered the one or more post-operative complications (Shelton [0354] “In an example, the surgical hub 20326 may be communicably coupled to one or more surgeon and/or patient sensing systems 20336. The sensing systems 20336 may be used to measure and/or monitor, in real-time, various biomarkers associated with a surgeon performing a surgical procedure or a patient on whom a surgical procedure is being performed. A list of the patient/surgeon biomarkers measured by the sensing systems 20336 is provided herein. In an example, the surgical hub 20326 may be communicably coupled to an environmental sensing system 20334. The environmental sensing systems 20334 may be used to measure and/or monitor, in real-time, environmental attributes, for example, temperature/humidity in the surgical theater, surgeon movements, ambient noise in the surgical theater caused by the surgeon's and/or the patient's breathing pattern, etc.” and Shelton [0373] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor. The processor may be configured to perform the method. A biomarker may be received for a patient from a sensing system. A data collection that includes pre-surgical data may be received for a patient. A probability of a patient outcome due to a surgery performed on the patient may be determined using the biomarker and the data collection. A notification may be sent to a user. The notification may indicate that the probability of the surgical complication may exceed a threshold.”);
generating pre-processed time-marked data by (Shelton [0354] “In an example, the surgical hub 20326 may be communicably coupled to one or more surgeon and/or patient sensing systems 20336. The sensing systems 20336 may be used to measure and/or monitor, in real-time, various biomarkers associated with a surgeon performing a surgical procedure or a patient on whom a surgical procedure is being performed. A list of the patient/surgeon biomarkers measured by the sensing systems 20336 is provided herein. In an example, the surgical hub 20326 may be communicably coupled to an environmental sensing system 20334. The environmental sensing systems 20334 may be used to measure and/or monitor, in real-time, environmental attributes, for example, temperature/humidity in the surgical theater, surgeon movements, ambient noise in the surgical theater caused by the surgeon's and/or the patient's breathing pattern, etc.” and Shelton [0366] “Data preparation may be performed for machine learning as another stage of the machine learning lifecycle. Data preparation may include data preprocessing steps such as data formatting, data cleaning, and data sampling.”:
filtering and augmenting, via a filtering and augmentation engine of the one or more processors, the time-marked data for each patient of the second plurality of patients by removing erroneous measurements from the time-marked data, removing the time-marked data after diagnosis of the one or more post-operative complications, and decimating the time-marked data into time-based means (Shelton [0402] “For example, the aggregation and filtering, 29010, may include filtering (e.g., to select specific sensor data from the stream of data from the pre-surgical data collection 29002). For example, the aggregation and filtering, 29010, may include averaging (e.g., to establish a baseline for a specific biomarker from the pre-surgical data collection 29002). For example, the aggregation and filtering, 29010, may include correlation analysis (e.g., to establish a baseline for relationships between and/or among specific biomarkers from the pre-surgical data collection 29002). For example, the aggregation and filtering, 29010, may include data translation (e.g., to coordinate format and/or datatype differences between a data source and the format and datatype expected by the contextual database 29024).”, and Shelton[0366] “Such identifying data may be removed before model training. For example, identifying data may be removed for privacy reasons. As another example, data may be removed because there may be more data available than may be used for model training. In such case, a subset of the available data may be randomly sampled and selected for model training and the remainder may be discarded.” and Shelton [0068] “The blood pressure sensing system may display blood pressure information locally or transmit the data to a system. The sensing system may display blood pressure information graphically over a period of time.”); and
standardizing, via the standardization engine of the one or more processors, the time-based means to have a mean of 0 and a standard deviation of 1 (Shelton [0367] “Data preparation may include data transforming procedures (e.g., after preprocessing), such as scaling and aggregation. For example, the preprocessed data may include data values in a mixture of scales. These values may be scaled up or down, for example, to be between 0 and 1 for model training. For example, the preprocessed data may include data values that carry more meaning when aggregated. In an example, there may be multiple prior colorectal procedures a patient has had. The total count of prior colorectal procedures may be more meaningful for training a model to predict surgical complications due to adhesions. In such case, the records of prior colorectal procedures may be aggregated into a total count for model training purposes.”);
inputting, by the one or more processors, the pre-processed time-marked data into the one or more real-time data risk assessment models (Shelton [0368] “Model training may be another aspect of the machine learning lifecycle. The model training process as described herein may be dependent on the machine learning algorithm used. A model may be deemed suitably trained after it has been trained, cross validated, and tested. Accordingly, the dataset from the data preparation stage (e.g., an input dataset) may be divided into a training dataset (e.g., 60% of the input dataset), a validation dataset (e.g., 20% of the input dataset), and a test dataset (e.g., 20% of the input dataset). After the model has been trained on the training dataset, the model may be run against the validation dataset to reduce overfitting. If accuracy of the model were to decrease when run against the validation dataset when accuracy of the model has been increasing, this may indicate a problem of overfitting. The test dataset may be used to test the accuracy of the final model to determine whether it is ready for deployment or more training may be required.”); and
performing, by the one or more processors, one or more regression analyses on the pre-processed time-marked data to determine a relationship between the pre-processed time-marked data and the one or more risk scores (Shelton [0364] “The output of machine learning's training process may be a model for predicting outcome(s) on a new dataset. For example, a linear regression learning algorithm may be a cost function that may minimize the prediction errors of a linear prediction function during the training process by adjusting the coefficients and constants of the linear prediction function. When a minimal may be reached, the linear prediction function with adjusted coefficients may be deemed trained and constitute the model the training process has produced.” And Shelton [0360] “Machine learning may be supervised (e.g., supervised learning). A supervised learning algorithm may create a mathematical model from training a dataset (e.g., training data). The training data may consist of a set of training examples. A training example may include one or more inputs and one or more labeled outputs. The labeled output(s) may serve as supervisory feedback. In a mathematical model, a training example may be represented by an array or vector, sometimes called a feature vector. The training data may be represented by row(s) of feature vectors, constituting a matrix. Through iterative optimization of an objective function (e.g., cost function), a supervised learning algorithm may learn a function (e.g., a prediction function) that may be used to predict the output associated with one or more new inputs. A suitably trained prediction function may determine the output for one or more inputs that may not have been a part of the training data. Example algorithms may include linear regression, logistic regression, and neutral network. Example problems solvable by supervised learning algorithms may include classification, regression problems, and the like.”, Shelton [0118] “Based on the detected circulating tumor cells data, the circulating tumor cell sensing system may calculate metastatic risk. The presence of circulating cancerous cells may indicate metastatic risk. Circulating cancerous cells per milliliter of blood exceeding a threshold amount may indicate a metastatic risk. Cancerous cells may circulate the bloodstream when tumors metastasize. Based on the calculated metastatic risk, the circulating tumor cell sensing system may generate a surgical risk score. Based on the generated surgical risk score, the circulating tumor cell sensing system may indicate surgery viability and/or suggested surgical precautions.”, Shelton [0354] “In an example, the surgical hub 20326 may be communicably coupled to one or more surgeon and/or patient sensing systems 20336. The sensing systems 20336 may be used to measure and/or monitor, in real-time, various biomarkers associated with a surgeon performing a surgical procedure or a patient on whom a surgical procedure is being performed. A list of the patient/surgeon biomarkers measured by the sensing systems 20336 is provided herein. In an example, the surgical hub 20326 may be communicably coupled to an environmental sensing system 20334. The environmental sensing systems 20334 may be used to measure and/or monitor, in real-time, environmental attributes, for example, temperature/humidity in the surgical theater, surgeon movements, ambient noise in the surgical theater caused by the surgeon's and/or the patient's breathing pattern, etc.” and Shelton [0005] “A computing system and/or a method may be provided for using a risk assessment to provide a notification. The computing system may comprise a processor. The processor may be configured to perform the method. A biomarker may be received for a patient from a sensing system. A data collection that includes pre-surgical data may be received for a patient. A probability of a patient outcome due to a surgery performed on the patient may be determined using the biomarker and the data collection. A notification may be sent to a user. The notification may indicate that the probability of the surgical complication may exceed a threshold.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to train the real time risk assessment models using preprocessed and standardized time marked sensor data, as Shelton teaches real time biosensor measurement, filtering sensor streams, removing erroneous or irrelevant values, aggregating data into time based means, scaling values, and training supervised regression models. Shelton provides motivation to prepare and model sensor data in this manner to assess surgical complications in real time. Accordingly, one of ordinary skill would have been motivated to adopt Shelton’s preprocessing techniques, including filtering noise, aggregating time based sensor measurements, standardizing numerical values, and training regression models on labeled complication outcomes, to develop a real time risk model, because these steps use conventional machine learning practices for time series biomedical data and would predictably yield improved complication detection performance.
Regarding claim 24, Shelton, Burnett, Haber, and Janssen teach the invention in claim 23, as discussed above, and further teach further comprising using the real-time data to further train the one or more real-time risk assessment models (Shelton [0450]“In an example, the previous patient outcomes and/or risk models may be trained. The trained risk models may provide more accurate data for analysis. Machine learning may be provided to train the previous patient outcomes and/or risk models. The machine learning may be supervised and/or unsupervised, for example. The machine learning may include updating the previous patient outcomes data and risk models with recent analyses. The machine learning's training process may update the risk model for predicting an outcome on new data set. For example, previous patient outcomes data and risk models may be updated with new analyses and determinations relating to risk of septic heart rate. The machine learning may use the determined risk of complication to train the risk model, for example. After an analysis, the previous patient outcomes and/or risk models may be updated to provide more accurate data for future analysis.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to update and retrain the real time risk assessment models using new real time data because Shelton teaches continuous model updating, retraining, and modifying biases and weights based on new patient outcomes, providing motivation to refine machine learning models using current sensor data.
Regarding claim 25, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach wherein the historical data comprises one or more of medical records, surgical history, individual health indicators, and surgical parameters associated with the first plurality of patients (Shelton [0396] “For example, the EMR 29007 may include patient medical records. The EMR 29007 may include any data source relevant to a patient in view of a health procedure. The patient medical records may include at least one of medical history of the patient, patient demographics, past procedures, medications, treatment plans, immunization dates, allergies, radiology images, laboratory and test results, notes, and the like, for example.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to include medical records, surgical history, health indicators, and surgical parameters in the historical dataset because Shelton teaches that EMR data includes medical histories, demographics, vitals, prior surgeries, medications, and lab results, motivating the inclusion of these exact categories in any risk assessment model.
Regarding claim 26, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach wherein the time-marked data comprises biomarker data measured by a plurality of sensors coupled to the second plurality of patients (Shelton [0039] “FIG. 1B is a block diagram of an example relationship among sensing systems 20001, biomarkers 20005, and physiologic systems 20007. The relationship may be employed in the computer-implemented patient and surgeon monitoring system 20000 and in the systems, devices, and methods disclosed herein. For example, the sensing systems 200001 may include the wearable sensing system 20011 (which may include one or more surgeon sensing systems and one or more patient sensing systems) and the environmental sensing system 20015 as discussed in FIG. 1A. The one or more sensing systems 20001 may measure data relating to various biomarkers 20005. The one or more sensing systems 20001 may measure the biomarkers 20005 using one or more sensors, for example, photosensors (e.g., photodiodes, photoresistors), mechanical sensors (e.g., motion sensors), acoustic sensors, electrical sensors, electrochemical sensors, thermoelectric sensors, infrared sensors, etc. The one or more sensors may measure the biomarkers 20005 as described herein using one of more of the following sensing technologies: photoplethysmography, electrocardiography, electroencephalography, colorimetry, impedimentary, potentiometry, amperometry, etc..”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to use biomarker data measured by sensors as part of the time marked dataset because Shelton teaches that sensing systems collect biomarker data using multiple sensor types , motivating the use of such sensor based measurements for time stamped postoperative monitoring.
Regarding claim 27, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach further comprising applying weighted coefficients to the one or more historical data risk assessment models and the one or more real-time risk assessment models based on a current condition of the target patient (Shelton [0371] “For example, a deployed model may be updated as more live production data become available as training data. In such case, the deployed model may be further trained, validated, and tested with such additional live production data. In an example, the updated biases and weights of a further-trained MLP model may update the deployed MLP model's biases and weights. Those skilled in the art recognize that post-deployment model updates may not be a one-time occurrence and may occur as frequently as suitable for improving the deployed model's accuracy.” and Shelton [0499] “For example, an analysis may determine that blood pressure levels are in a normal range. The analysis may determine that the oxygen saturation levels are in a normal range. The analysis may determine that the heart rate levels are in an abnormal range. The analysis may generate a risk level based on the weighted biomarker analysis. The risk level may indicate a moderate risk. The risk level may indicate a moderate risk based on the weighted biomarker analysis.”, Shelton [0444] “FIG. 15 depicts a block diagram of a system for analyzing one or more biomarkers using machine learning and a data collection. At 29043, a first biomarker may be measured, such as heart rate, for example. The measured biomarker may indicate a risk of complication, such as septic heart rate, for example. At 29044, one or more biomarkers associated with the first biomarker and the complication may be determined. At 29045, previous patient outcome data and/or risk models based on a risk may be retrieved. At 29046, an analysis may be performed. At 29047, the risk of complication may be determined, such as probability of septic heart rate, for example. An output may be generated based on the risk of complication. At 29048, the previous patient outcome data and/or risk models may be updated.” and Shelton [0445] “In an example, a heart rate may be monitored. The heart rate may be monitored by a sensing system and/or wearable, for example. At 29043, based on the heart rate monitoring, an abnormal heart rate may be detected, such as a high heart rate, for example. A risk of a complication, such as septic heart rate, may be indicated. Risk of septic heart rate may be determined based on the heart rate measurement.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to apply weighted coefficients to the historical and real time risk assessment models based on the current patient condition because Shelton teaches updating weights and biases of machine learning models based on new physiologic data, and performing weighted biomarker analyses to generate risk levels, providing clear motivation to weight risk factors dynamically.
Regarding claim 28, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach further comprising notifying a user of the risk assessment of the target patient (Shelton [0005] “A computing system and/or a method may be provided for using a risk assessment to provide a notification.”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to incorporate a user notification step into the method, as taught by Shelton, because Shelton teaches that risk assessments are used to provide notifications to users, thereby motivating the combination and providing a predictable result, informing clinicians or patients when a complication risk is detected.
Regarding claim 29, Shelton, Burnett, Haber, and Janssen teach the invention in claim 28, as discussed above, and further teach wherein the notifying is triggered once a series of consecutive post-operative time intervals linked to the high risk score exceeds a predetermined limit (Shelton [0356] “In an example, the surgical hub 20326 may compare the measurement data from the sensing systems 20336 with one or more thresholds defined based on baseline values, pre-surgical measurement data, and/or in surgical measurement data. The surgical hub 20326 may compare the measurement data from the sensing systems 20336 with one or more thresholds in real-time. The surgical hub 20326 may generate a notification for displaying. The surgical hub 20326 may send the notification for delivery to a human interface system for patient 20339 and/or the human interface system for a surgeon or an HCP 20340, for example, if the measurement data crosses (e.g., is greater than or lower than) the defined threshold value. The determination whether the notification would be sent to one or more of the to the human interface system for patient 20339 and/or the human interface system for an HCP 2340 may be based on a severity level associated with the notification. The surgical hub 20326 may also generate a severity level associated with the notification for displaying”).
Therefore, it would be obvious to a PHOSITA before the effective filing date of the invention to trigger the notification based on a defined number of consecutive high risk intervals, because Shelton teaches repeatedly comparing real time sensor data to threshold criteria and generating notifications when the threshold is crossed, including severity based rules for alerting. A PHOSITA would recognize that evaluating multiple consecutive intervals is a predictable variation of the same threshold based notification logic to reduce false positives.
Claims 30-39 are analogous to claims 20-29, thus claims 30-39 are similarly analyzed and rejected in a manner consistent with the rejection of claims 20-29.
Regarding claim 40, Shelton, Burnett, Haber, and Janssen teach the invention in claim 20, as discussed above, and further teach wherein the temporal filter is selected from a group consisting of a cumulative average, a moving average filter, a gaussian window filter, or an exponential filter (Janssen [0033] “FIG. 5 is another representative image showing the monitored data obtained from a patient over the same four-day monitoring and analysis period. The trace 46, which is an ECG artifact-based example, represents an exemplary time series smoothing approach with a twenty-minute moving average of the total alert frequency/count while trace 47 is a sixty-minute moving average of the total time in alert shown over the four-day monitoring and analysis period. In the monitor alert interpretation display 48 of FIG. 5, a determined artifact-specific enhanced threshold 50a is shown for the twenty-minute moving average, which is set at sixty-five alerts per twenty minute threshold. The enhanced threshold 50b is shown for the sixty-minute moving average of the total time in alert and is set at 2.5 minutes of artifact alert per the sixty minute threshold. In addition to the traces 46 and 47, lead failures 51 are shown in the upper portion 49 of the display. In the time periods shown by reference numeral 52 and reference numeral 61, a significant number of artifacts (i.e., over sixty-five artifact alerts per twenty minutes in the illustrated example) are detected and may be indicated to the monitoring technician each time they occur. The present disclosure addresses this issue by providing a contextual enhanced alert or “meta-alert” and supporting visual aids identifying these sustained alert cases for each parameter or combinations of parameters. In the embodiment shown in FIG. 5, the enhanced threshold 50a is set for the moving average of the number of artifacts that occur over a selected period of time. As an illustrative example, the enhanced threshold 50a could be set at an intensity of sixty-five occurrences over a twenty minute period. It should be understood that the enhanced threshold 50a is set separately for each type of alert being monitored and that the intensity could be changed/modified by each facility. When the tracked moving average exceeds the enhanced threshold 50a, an enhanced alert is generated. Such enhanced alert will allow the monitoring technician who is managing multiple patients to recognize the repeated generation of individual artifact alerts, understand the longitudinal context of the frequency of the alerts, individual alert duration, the accumulated duration of the alerts, and the potential repeated and sustained nature of the alerts. Without such an enhanced alert, it is very difficult for the monitoring technician to grasp the frequency and duration of the artifact alerts.”).
It would have been obvious to a PHOSITA before the effective filing date of the invention to utilize a temporal filter, including a cumulative average, moving average filter, gaussian window filter, or exponential filter, in the postoperative physiological monitoring and risk assessment system of Shelton, Burnett, Haber, and Janssen because Janssen teaches applying moving average and time-series smoothing techniques to physiological monitoring data over defined monitoring intervals in order to evaluate sustained physiological conditions, reduce transient noise, analyze accumulated alert frequency and duration, and improve the reliability of generated alerts. A PHOSITA would have recognized that this temporal filtering techniques were well-known and commonly used to smooth and improve the stability and accuracy of threshold monitoring systems, and would therefore have been motivated to implement one or more known temporal filters, including moving average, cumulative average, gaussian window, or exponential filtering techniques, within the combined postoperative risk assessment framework to improve robustness in data, yielding predictable results consistent with the established functions of these filters.
Claim 41 is analogous to claim 40, thus claim 41 is similarly analyzed and rejected in a manner consistent with the rejection of claim 40.
Response to Arguments
Applicant’s arguments and amendments, see Remarks/Amendments submitted on 02/12/2026 with respect to the rejection of the claims have been carefully considered and is addressed below.
Claim Rejections - 35 USC § 112
The amendment to claims 20 and 30 clarifies the relationship between the first threshold, second threshold, and cumulative metric, including how each value is determined and compared to generate a risk score. Accordingly, the indefiniteness issue previously identified under 35 U.S.C. § 112(b) has been overcome, and the rejection is withdrawn.
Claim Rejections - 35 USC § 103
Applicant’s arguments have been fully considered but are not persuasive. The rejection is based on the combined teachings of Shelton, Burnett, Haber, and Janssen, each of which addresses complementary aspects of physiological monitoring, predictive modeling, threshold generation, and temporal analysis. Shelton teaches risk assessment models trained using patient datasets, generation of complication probabilities and surgical risk scores, use of historical and real-time biomarker information, and continuous postoperative monitoring using cloud-based processing. Shelton further teaches model training, updating of risk models using previous patient outcomes, and real-time monitoring of biomarkers before and after surgery. Accordingly, Shelton reasonably teaches the use of historical patient information and continuously monitored physiological data within predictive risk assessment models.
Applicant’s arguments regarding the claimed thresholding and temporal filtering limitations are not unpersuasive because the rejection relies on the combined teachings of Haber and Janssen, not Shelton alone, for these features. Haber teaches generating alert thresholds using outcome and non-outcome patient datasets and ROC threshold selection while accounting for “time to risk” during predictive modeling. Janssen further teaches analyzing physiological monitoring data over defined monitoring intervals using moving averages, cumulative alert frequency analysis, enhanced thresholds, and smoothing algorithms applied to continuously monitored time-series data. Janssen discloses determining thresholds over selected monitoring periods, evaluating accumulated duration and frequency of alerts over time, and generating enhanced alerts when cumulative metrics exceed threshold values. Collectively, these teachings reasonably suggest determining thresholds across monitoring intervals and applying temporal filtering to physiological time-series data in order to improve alert reliability and reduce noise.
Applicant additionally states that Burnett’s inline sensor device is not a predictable alternative to Shelton’s disclosed sensing systems. However, Burnett teaches inline Foley catheter sensing systems configured to continuously measure biomarker data from bodily fluids flowing through a fluid channel and transmit the resulting physiological measurements to remote controllers and servers for analysis. Shelton teaches integrating sensing systems and biomarker measurements into postoperative complication prediction workflows. A PHOSITA would have recognized Burnett’s inline sensing system as a compatible physiological sensing mechanism capable of supplying additional real-time biomarker data to Shelton’s predictive monitoring architecture. The combination substitutes one known physiological sensing modality for another to improve the specificity, continuity, and accuracy of biomarker acquisition, which represents the predictable use of prior art elements according to their established functions.
Conclusion
The prior art made of record and not relied upon is considered pertinent to Applicant's disclosure.
Itu et al. (U.S. Publication 2021/0251577 A1) teaches a machine learning system that predicts and reduces peri-procedural complications, such as myocardial infarction during or around PCI procedures, by analyzing combinations of imaging and non-imaging data before, during, or after the procedure to provide risk assessment and treatment recommendations.
Shelton et al. (U.S. Publication 2022/0233254 A1) teaches a computing system that uses pre and intra operative biomarker measurements to predict hemostasis related complications and based on that prediction, generate control signals or recommendation to adjust surgical parameters, device operations, procedure plans, or instrument selection.
Lu et al. (CN Patent Publication 113178258 A) teaches a preoperative risk assessment method and system that uses historical patient data, feature engineering, and supervised learning to predict the likelihood of surgical complications and generate a risk assessment report, thereby improving the reliability of complication risk evaluation.
Wennberg (U.S. Publication 2006/0129428 A1) teaches a system for predicting healthcare financial risk by assessing and filtering patient, geographic, and healthcare system data into clean data, then applying a predictive model to generate patient profiles and identify those at risk.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KYRA R LAGOY whose telephone number is (703)756-1773. The examiner can normally be reached Monday - Friday, 8:00 am - 5:00 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kambiz Abdi can be reached at (571)272-6702. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/K.R.L./Examiner, Art Unit 3685
/KAMBIZ ABDI/Supervisory Patent Examiner, Art Unit 3685
Prosecution Insights