DETAILED ACTION
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention. Independent Claims 1, 10 similarly recite limitation to perform the functions including “inputting the medical image into a machine learning model, wherein the machine learning model is configured to determine a quality value based on the medical image”, however, what is “a machine learning model”, and what is “a quality value” are not defined, so that it is not clear what, and how “quality” is determined, and what “quality”, or “quality value” associated with “medical image” are being applied and determined, and in order to determine the “quality”, what are the parameters, or measurements, or features, or attributions….etc., are being evaluated?
Furthermore, it is not clear whether or not above “a machine learning model” is a trained model or not; dependent claim 4 recites limitation of “ wherein when training the second machine learning model”, and, dependent claim 9 recites limitation “wherein when training the machine learning model,”, which lacks antecedent basis, because it is not mentioned in claim 1 about “training the machine learning model”, neither as pointed above, “a machine learning model” recited in claim 1 is a trained model or not.
Furthermore, claim 4 recites “labeling ….an image feature associated with the target clinical application”, since in claim 1, what is “a target clinical application”, or at least the scope of “a target clinical application” is not defined, thus, “one or more features, associated with the target clinical application”, and, “an image feature associated with the target clinical application”, which is recited in claims 1, 3, and in claim 4, is not defined, the scope of “an image feature associated with the target clinical application” is not defined, such as, is this “an image feature associated with the target clinical application”, which is recited in claims 1, 3, and in claim 4 are the same image feature (s), or not; and it is well known that there can be a very different, multiple variations in features, and/or image features and clinical applications, it is indefinite in the scope of the claim limitations as required particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Thus, overall, these render claims 1-20 indefinite.
Claim Rejections - 35 USC § 103
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 of this title, 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.
Claims 1-6, 10-16, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over LIU (US 20200049785 A1, as provided as IDS), and in view of INAM (WO 2022011342 A1, claims priority date of US Provisional Application 63/050,333, July 10, 2020, as provided as IDS).
Re Claim 1, LIU discloses a method for assessing medical images for suitability in clinical interpretation (see LIU: e.g., Fig. 1, -- A scoring system assesses quality and helps determine whether enough significant clinical value may be extracted and therefore lead to correct diagnosis--, in [0002]); the method comprising:
acquiring a medical image of an anatomical region from a medical imaging device, at a first location (see LIU: e.g., Fig. 1, -- A scoring system assesses quality and helps determine whether enough significant clinical value may be extracted and therefore lead to correct diagnosis--, in [0002], and, --machine-based deep learning to train a network to predict motion artifacts given an input 3D representation from a scan of a patient. The architecture of the network may be defined to deal with anisotropic data from the MR scan. [0005] In a first aspect, a method is provided for machine learning to determine motion artifacts for a magnetic resonance system. The magnetic resonance system generates a first three-dimensional representation of a patient.--, in [0003]-[0005]; and, --deep learning is provided for volumetric MR images. The motion severity is assed for volumetric MR images (e.g., multi-slice or full 3D) using deep learning. The quality of an MR scan is evaluated based on the whole volume rather than a 2D image from the volume--, in [0015], and [0023]);
inputting the medical image into a machine learning model, wherein the machine learning model is configured to determine a quality value based on the medical image (see LIU: e.g., Fig. 1, -- A scoring system assesses quality and helps determine whether enough significant clinical value may be extracted and therefore lead to correct diagnosis--, in [0002], and, --machine-based deep learning to train a network to predict motion artifacts given an input 3D representation from a scan of a patient. The architecture of the network may be defined to deal with anisotropic data from the MR scan. [0005] In a first aspect, a method is provided for machine learning to determine motion artifacts for a magnetic resonance system. The magnetic resonance system generates a first three-dimensional representation of a patient.--, in [0003]-[0005]; and, --Once the quality score for motion artifact in an MR image volume is predicted, the operator of the medical scanner or the medical scanner decides whether to rescan the patient. The score is used for a decision to use or not use the generated 3D representation. The result is that a later physician review is more likely to have a useful image for diagnosis, and rescanning is avoided where possible. The score may be used to weight an amount of trust in diagnosis based on a 3D representation reconstructed from a scan of the patient.--, in [0073]);
determining whether the quality value meets a quality threshold associated with a target clinical application (see LIU: e.g., --[0073] The user or the medical scanner uses the quality score. A sufficiently good quality 3D representation (e.g., score or value above or below a threshold) allows for diagnosis with lest risk for error. A poor-quality 3D representation due to the combination of patient motion with the line order may not be sufficient for diagnosis,--, in [0073]);
and, on the basis of at least meeting the quality threshold associated with the target clinical application storing the medical image at a storage location (see Liu: e.g., -- the machine-learned network includes one or more layers with values for various parameters, such as convolution kernels, down sampling weights, and/or connections. The values of the parameters and/or the network as trained are stored in act 19. The machine-learned network is stored in a memory, such as memory of the machine or the database with the examples. The machine-learned network may be transmitted to a different memory. The machine-learned neural network may be duplicated for application by other devices or machines, such as processors of MR scanners. The memories of MR scanners may store copies of the machine-learned network for application for specific patients, enabling a radiologist or other physician to determine whether to rely on an image or to scan again for diagnosis due to patient motion. [0060] FIG. 6 is a flow chart diagram of one embodiment of a method for determining motion artifact for a magnetic resonance system. The stored machine-learned network is applied to determine a score for a scan of a patient. A 3D scan of the patient is performed, and the level of motion artifact from the resulting 3D representation of that patient is determined by the machine-learned network.—in [0059]-[0060]; and, -- The quality score is transmitted over a network, through a communications interface, into memory or database (e.g., to a computerized patient medical record),--, in [0069]-[0070]; and, -- FIG. 7 shows one embodiment of a system for machine learning and/or for application of a machine-learned network. The system is distributed between the imaging system 80 and a remote server 88. In other embodiments, the system is just the server 88--, in [0074]-[0075]);
LIU however does not explicitly disclose accessing the medical image from the storage location, said accessing being from a second location, different from the first location, and analyzing the medical image at the second location, to identify one or more features therein, associated with the target clinical application,
INAM disclose accessing the medical image from the storage location, said accessing being from a second location, different from the first location, and analyzing the medical image at the second location, to identify one or more features therein, associated with the target clinical application (see INAM: e.g., -- Analyzing the input radiographic image data may include detecting anomalous features in the input radiographic image data, including determining one or more of, for example, whether a portion of the input radiographic image data substantially matches a portion of a previously stored radiographic image, and/or whether a portion of the input radiographic image data was modified…. by one or more machine learning models implemented by the computing system, the input radiographic image data to identify one or more dental features associated with the at least one dental object, and derive, by the computing system, based on the treatment data and the identified one or more dental features associated with the at least one dental object, one or more integrity scores for the input radiographic image data and the treatment data, with the one or more integrity scores being representative of potential integrity problems associated with the input radiographic image data and the treatment data. --, in [0040]-0041]);
LIU and INAM are combinable as they are in the same field of endeavor: to analysis of medical images and determination of image quality based on machine learning models. Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify LIU’s method using INAM’s teachings by including accessing the medical image from the storage location, said accessing being from a second location, different from the first location, and analyzing the medical image at the second location, to identify one or more features therein, associated with the target clinical application to LIU’s analysis of the medical image {see LIU’s: -- The medical image or dataset is acquired by the medical MR scanner. Alternatively, the acquisition is from storage or memory, such as acquiring a previously created dataset from a PACS. A processor may extract the data from a picture archive communications system or a medical records database. Acquisition may be through transmission over a network--, in [0023], and, -- [0052] The network is trained with training data. Samples of input data with ground truth are used to learn to classify the score. For deep learning, the classifier learns the features of the input data to extract from the training data. Alternatively, the features, at least for the input, are manually programmed, such as filtering the scan data and inputting the results of the filtering. --, in [0052]} in order to provide integrity scores for the input radiographic image data and the treatment data for the target clinical application, such as diagnosis and treatment (see INAM: e.g., in [0040]-[0041]).
Re Claim 2, LIU as modified by INAM further disclose wherein the method is performed on a computing device, the storage location is a server and prior to the analyzing, the method further comprises: transmitting, from the computing device to the server, the medical image and the quality value of the machine learning model; storing the medical image and the quality value on the server; and accessing, from the server, the medical image (see Liu: e.g., -- the machine-learned network includes one or more layers with values for various parameters, such as convolution kernels, down sampling weights, and/or connections. The values of the parameters and/or the network as trained are stored in act 19. The machine-learned network is stored in a memory, such as memory of the machine or the database with the examples. The machine-learned network may be transmitted to a different memory. The machine-learned neural network may be duplicated for application by other devices or machines, such as processors of MR scanners. The memories of MR scanners may store copies of the machine-learned network for application for specific patients, enabling a radiologist or other physician to determine whether to rely on an image or to scan again for diagnosis due to patient motion. [0060] FIG. 6 is a flow chart diagram of one embodiment of a method for determining motion artifact for a magnetic resonance system. The stored machine-learned network is applied to determine a score for a scan of a patient. A 3D scan of the patient is performed, and the level of motion artifact from the resulting 3D representation of that patient is determined by the machine-learned network.—in [0059]-[0060]; and, -- The quality score is transmitted over a network, through a communications interface, into memory or database (e.g., to a computerized patient medical record),--, in [0069]-[0070]; and, -- FIG. 7 shows one embodiment of a system for machine learning and/or for application of a machine-learned network. The system is distributed between the imaging system 80 and a remote server 88. In other embodiments, the system is just the server 88--, in [0074]-[0075]).
Re Claim 3, LIU as modified by INAM further disclose wherein the machine learning model is a first machine learning model, and the analyzing the medical image at the second location comprises:
inputting the medical image into a second machine learning model, wherein the second machine learning model is configured to detect the one or more features therein, associated with the target clinical application, for medical images that meet the quality threshold associated with the target clinical application (see INAM: e.g., -- Analyzing the input radiographic image data may include detecting anomalous features in the input radiographic image data, including determining one or more of, for example, whether a portion of the input radiographic image data substantially matches a portion of a previously stored radiographic image, and/or whether a portion of the input radiographic image data was modified…. by one or more machine learning models implemented by the computing system, the input radiographic image data to identify one or more dental features associated with the at least one dental object, and derive, by the computing system, based on the treatment data and the identified one or more dental features associated with the at least one dental object, one or more integrity scores for the input radiographic image data and the treatment data, with the one or more integrity scores being representative of potential integrity problems associated with the input radiographic image data and the treatment data. --, in [0040]-0041]).
Re Claim 4, LIU as modified by INAM further disclose wherein when training the second machine learning model, the method further comprises:
accessing a set of training medical images (see INAM: e., -- a system can be trained using data generated by a ML model in order to identify patterns in dental radiographs. Such determinations may be made following the accumulation, review, and/or analysis of user data from a large number of users over time as well as individual patient data, that may be configured to provide the proposed ML approaches with an initial or ongoing training set. In addition, in some implementations, supplemental training data may be intermittently provided to fine-tune or increase the effectiveness of the machine learning model implemented by the machine learning system. [083] In different implementations, a training system may be used that includes an initial ML model (which may be referred to as an “ML model trainer”) configured to generate a subsequent trained ML model from training data obtained from a training data repository or from device-generated data. The generation of this ML model may be referred to as “training” or “learning.” The training system may include and/or have access to substantial computation resources for training, such as a cloud, including computer server systems adapted for machine learning training. In some implementations, the ML model trainer is configured to automatically generate multiple different ML models from the same or similar training data for comparison.--, in [0082]-[0083]);
labelling, for each of the set of training medical images, an image feature associated with the target clinical application (see INAM: e.g., --the training system (whether stored remotely, locally, or both) can be configured to receive and accumulate more and more training data items, thereby increasing the amount and variety of training data available for ML model training, resulting in increased accuracy, effectiveness, and robustness of trained ML models. [085] FIG.1A illustrates an example system 100 upon which aspects of this disclosure may be implemented. The system 100 may include a computing platform (e.g., a personal computer) 105 executing software that implements an image processor 200 (e.g., implemented as a ML model) which is coupled to receive a radiographic image 120 containing an oral structure 125 that could comprise a natural tooth, an implant, a restoration, etc., that further relates to craniofacial structures such as bones. The image processor 200 provides output, such as image metadata (e.g., detected features that appear in the image, and their corresponding labels) to a measurement processor 300 which in turn provides calibrated output (a calibrated measurement) based on the image metadata and data from a library 150 that may include an X-ray sensor database, an implant database, a population based anatomical averages database, and a patient specific database, as more particularly described below. [086] A training mechanism 140 provides a machine-learning based training mechanism, as mentioned above, for training aspects of the image processor 200 for generating (predicting), masks, labels, features and points of the oral structure--, in [0083]-[0088]); and
training the second machine learning model to identify the image feature on future medical images (see INAM: e.g., -- a system can be trained using data generated by a ML model in order to identify patterns in dental radiographs. Such determinations may be made following the accumulation, review, and/or analysis of user data from a large number of users over time as well as individual patient data, that may be configured to provide the proposed ML approaches with an initial or ongoing training set. In addition, in some implementations, supplemental training data may be intermittently provided to fine-tune or increase the effectiveness of the machine learning model implemented by the machine learning system. [083] In different implementations, a training system may be used that includes an initial ML model (which may be referred to as an “ML model trainer”) configured to generate a subsequent trained ML model from training data obtained from a training data repository or from device-generated data. The generation of this ML model may be referred to as “training” or “learning.” The training system may include and/or have access to substantial computation resources for training, such as a cloud, including computer server systems adapted for machine learning training. In some implementations, the ML model trainer is configured to automatically generate multiple different ML models from the same or similar training data for comparison.--, in [0082]-[0083]).
Re Claim 5, LIU as modified by INAM further disclose wherein to determine the quality threshold associated with the target clinical application, the method further comprises: determining image quality values of the medical images in the set of training medical images; and setting the quality threshold associated with the target clinical application to correspond to the image quality values of the medical images in the set of training medical images used to train the second machine learning model (see LIU: e.g., --A machine trains a neural network to receive a second magnetic resonance reconstruction from a magnetic resonance scan and to output a score. The training uses the error and the training data. The machine-learned neural network is stored.--, in [0007], see INAM: e.g., -- threshold as desired using ML as provided by the training mechanism 140. The machine learning model may be based on specific patient characteristics as well as crowdsourced methods that provide for overall measurement techniques.--, in [0095]; also see: -- the dental data comprises input radiographic image data for at least one dental object, and identify, by one or more machine learning models, at least one first dental feature in the input radiographic image data for the at least one dental object, and at least one other feature in the dental object comprising at least partly a healthy dental structure. The computer readable media include further instructions to compute at least one dimensioned property representative of physical dimensions of the at least one first dental feature and the at least one other feature comprising at least partly the healthy dental structure, derive based on the at least one dimensioned property at least one dimensioned property ratio indicative of an extent of a dental clinical condition associated with the identified at least one first dental feature of the at least one dental object, and determine a treatment plan based on a comparison of the derived at least one dimensioned property ratio to a respective at least one pre-determined threshold value.--, in [0031]-0032], and, --. the generating a radiographic-based metric can be based on identifying anomalies within the data. The review metric can be based on a value, a range of values, a percentage, a threshold, and so on. The metric can be text based, where the text could include “pass” or “fail”, “good” or “bad”, “OK” or “Further examination needed”,--, in [0165]-[0166]).
Re Claim 6, LIU as modified by INAM further disclose determining whether the quality value of the medical image meets another quality threshold associated with another target clinical application; and when the quality value meets the another quality threshold associated with the another target clinical application, permitting storage of the medical image at the storage location and, at the second location, clinical interpretation of the medical image for the another target clinical application (see LIU: e.g., --A machine trains a neural network to receive a second magnetic resonance reconstruction from a magnetic resonance scan and to output a score. The training uses the error and the training data. The machine-learned neural network is stored.--, in [0007], see INAM: e.g., -- threshold as desired using ML as provided by the training mechanism 140. The machine learning model may be based on specific patient characteristics as well as crowdsourced methods that provide for overall measurement techniques.--, in [0095]; also see: -- the dental data comprises input radiographic image data for at least one dental object, and identify, by one or more machine learning models, at least one first dental feature in the input radiographic image data for the at least one dental object, and at least one other feature in the dental object comprising at least partly a healthy dental structure. The computer readable media include further instructions to compute at least one dimensioned property representative of physical dimensions of the at least one first dental feature and the at least one other feature comprising at least partly the healthy dental structure, derive based on the at least one dimensioned property at least one dimensioned property ratio indicative of an extent of a dental clinical condition associated with the identified at least one first dental feature of the at least one dental object, and determine a treatment plan based on a comparison of the derived at least one dimensioned property ratio to a respective at least one pre-determined threshold value.--, in [0031]-0032], and, --. the generating a radiographic-based metric can be based on identifying anomalies within the data. The review metric can be based on a value, a range of values, a percentage, a threshold, and so on. The metric can be text based, where the text could include “pass” or “fail”, “good” or “bad”, “OK” or “Further examination needed”,--, in [0165]-[0166]).
Re Claims 10-16, claims 10-16 are corresponding system claim to claims 1-6, respectively. Claims 10-16 thus are rejected for the similar reasons for claims 1-6. See above discussions with regard to claims 1-6 respectively. LIU as modified by INAM further disclose a system for assessing medical images for suitability in clinical interpretation, the system comprising: a first computing device, at a first location, comprising one or more device processors and a device memory storing device instructions for execution by the one or more device processors, wherein when the device instructions are executed by the one or more device processors, the one or more device processors are configured to perform the method (see LIU: e.g., Fig. 1, -- A scoring system assesses quality and helps determine whether enough significant clinical value may be extracted and therefore lead to correct diagnosis--, in [0002], and, --machine-based deep learning to train a network to predict motion artifacts given an input 3D representation from a scan of a patient. The architecture of the network may be defined to deal with anisotropic data from the MR scan. [0005] In a first aspect, a method is provided for machine learning to determine motion artifacts for a magnetic resonance system. The magnetic resonance system generates a first three-dimensional representation of a patient.--, in [0003]-[0005]; and, --Once the quality score for motion artifact in an MR image volume is predicted, the operator of the medical scanner or the medical scanner decides whether to rescan the patient. The score is used for a decision to use or not use the generated 3D representation. The result is that a later physician review is more likely to have a useful image for diagnosis, and rescanning is avoided where possible. The score may be used to weight an amount of trust in diagnosis based on a 3D representation reconstructed from a scan of the patient.--, in [0073]); and
a second computing device, at a second location, comprising one or more second device processors and a second device memory storing second device instructions for execution by the one or more second device processors, wherein when the second device instructions are executed by the one or more second device processors, the one or more second device processors are configured to perform the method (see Liu: e.g., -- the machine-learned network includes one or more layers with values for various parameters, such as convolution kernels, down sampling weights, and/or connections. The values of the parameters and/or the network as trained are stored in act 19. The machine-learned network is stored in a memory, such as memory of the machine or the database with the examples. The machine-learned network may be transmitted to a different memory. The machine-learned neural network may be duplicated for application by other devices or machines, such as processors of MR scanners. The memories of MR scanners may store copies of the machine-learned network for application for specific patients, enabling a radiologist or other physician to determine whether to rely on an image or to scan again for diagnosis due to patient motion. [0060] FIG. 6 is a flow chart diagram of one embodiment of a method for determining motion artifact for a magnetic resonance system. The stored machine-learned network is applied to determine a score for a scan of a patient. A 3D scan of the patient is performed, and the level of motion artifact from the resulting 3D representation of that patient is determined by the machine-learned network.—in [0059]-[0060]; and, -- The quality score is transmitted over a network, through a communications interface, into memory or database (e.g., to a computerized patient medical record),--, in [0069]-[0070]; and, -- FIG. 7 shows one embodiment of a system for machine learning and/or for application of a machine-learned network. The system is distributed between the imaging system 80 and a remote server 88. In other embodiments, the system is just the server 88--, in [0074]-[0075]).
Re Claims 19-20, claims 19-20 are corresponding system claim to claims 3, and 6, respectively. Claims 19-20 thus are rejected for the similar reasons for claims 3, 6. See above discussions with regard to claims 3, 6 respectively. LIU as modified by INAM further disclose wherein the machine learning model at the first computing device is a first machine learning model, and the analyzing of the medical image is at the storage device, which is a server, and comprises: inputting the medical image into a storage device machine learning model, wherein the storage device machine learning model is configured to detect the image feature associated with the target clinical application, for medical images that meet the quality threshold associated with the target clinical application (see LIU: e.g., Fig. 1, -- A scoring system assesses quality and helps determine whether enough significant clinical value may be extracted and therefore lead to correct diagnosis--, in [0002], and, --machine-based deep learning to train a network to predict motion artifacts given an input 3D representation from a scan of a patient. The architecture of the network may be defined to deal with anisotropic data from the MR scan. [0005] In a first aspect, a method is provided for machine learning to determine motion artifacts for a magnetic resonance system. The magnetic resonance system generates a first three-dimensional representation of a patient.--, in [0003]-[0005]; and, --Once the quality score for motion artifact in an MR image volume is predicted, the operator of the medical scanner or the medical scanner decides whether to rescan the patient. The score is used for a decision to use or not use the generated 3D representation. The result is that a later physician review is more likely to have a useful image for diagnosis, and rescanning is avoided where possible. The score may be used to weight an amount of trust in diagnosis based on a 3D representation reconstructed from a scan of the patient.--, in [0073]); and
wherein the storage device computing device is further configured to: determine whether the quality value meets another quality threshold associated with another target clinical application; and when the quality value meets the another quality threshold associated with the another target clinical application, permit clinical interpretation of the medical image for the another target clinical application (see Liu: e.g., -- the machine-learned network includes one or more layers with values for various parameters, such as convolution kernels, down sampling weights, and/or connections. The values of the parameters and/or the network as trained are stored in act 19. The machine-learned network is stored in a memory, such as memory of the machine or the database with the examples. The machine-learned network may be transmitted to a different memory. The machine-learned neural network may be duplicated for application by other devices or machines, such as processors of MR scanners. The memories of MR scanners may store copies of the machine-learned network for application for specific patients, enabling a radiologist or other physician to determine whether to rely on an image or to scan again for diagnosis due to patient motion. [0060] FIG. 6 is a flow chart diagram of one embodiment of a method for determining motion artifact for a magnetic resonance system. The stored machine-learned network is applied to determine a score for a scan of a patient. A 3D scan of the patient is performed, and the level of motion artifact from the resulting 3D representation of that patient is determined by the machine-learned network.—in [0059]-[0060]; and, -- The quality score is transmitted over a network, through a communications interface, into memory or database (e.g., to a computerized patient medical record),--, in [0069]-[0070]; and, -- FIG. 7 shows one embodiment of a system for machine learning and/or for application of a machine-learned network. The system is distributed between the imaging system 80 and a remote server 88. In other embodiments, the system is just the server 88--, in [0074]-[0075]; also see INAM: e.g., -- threshold as desired using ML as provided by the training mechanism 140. The machine learning model may be based on specific patient characteristics as well as crowdsourced methods that provide for overall measurement techniques.--, in [0095]; also see: -- the dental data comprises input radiographic image data for at least one dental object, and identify, by one or more machine learning models, at least one first dental feature in the input radiographic image data for the at least one dental object, and at least one other feature in the dental object comprising at least partly a healthy dental structure. The computer readable media include further instructions to compute at least one dimensioned property representative of physical dimensions of the at least one first dental feature and the at least one other feature comprising at least partly the healthy dental structure, derive based on the at least one dimensioned property at least one dimensioned property ratio indicative of an extent of a dental clinical condition associated with the identified at least one first dental feature of the at least one dental object, and determine a treatment plan based on a comparison of the derived at least one dimensioned property ratio to a respective at least one pre-determined threshold value.--, in [0031]-0032], and, --. the generating a radiographic-based metric can be based on identifying anomalies within the data. The review metric can be based on a value, a range of values, a percentage, a threshold, and so on. The metric can be text based, where the text could include “pass” or “fail”, “good” or “bad”, “OK” or “Further examination needed”,--, in [0165]-[0166]).
Claims 7-9, and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over LIU as modified by INAM, and in view of シェイ(JP 7051849 B2, claims priority of publication JP 2020500378, January 9, 2020, as provided as IDS).
Re Claim 7, LIU as modified by INAM disclose that related medical images can be IMR images of “[0026] The medical image represents tissue and/or bone structure of the patient. Alternatively, the medical image represents flow, velocity, or fluids within the patient. In other embodiments, the medical image represents both flow and structure.”,
LIU as modified by INAM however do not explicitly disclose identifying kidney stones, and the another target clinical application comprises identifying gallstones, and wherein the quality threshold associated with the target clinical application is higher than the quality threshold associated with the another target clinical application,
シェイ discloses identifying kidney stones, and the another target clinical application comprises identifying gallstones, and wherein the quality threshold associated with the target clinical application is higher than the quality threshold associated with the another target clinical application (see シェイ: e.g., -- Sources of variation are noise (eg mA, kVp, patient size, etc.), resolution (eg, reconstructed kernel type, thickness, pixel size, etc.), structured noise (eg, streaks, patterns, textures, etc.), shadows. Ing artifacts (eg, ribs, spinal reconstruction artifacts, etc.) can be included. The effect on the result can include measurement error, stage error, and the like. Another use case can include volume / size quantification of other organs (eg, kidney transplants, etc.) or masses within organs (eg, cysts or stones, etc.).--, in [0219] of JP 7051849 B2, and, --Similarly, in block 3116, the processed image data is analyzed. For example, image data is analyzed by DDLD1532 for quality, IQI, data quality index, other image quality metrics and the like. The DDLD 1532 learns from the content of machine-processable image data (eg, identified features, resolutions, noise, etc.) and reconstructs previous image data (eg, for the same patient 1406, same type, etc.) and / or reconstruction. Compare with the image. At block 3118, the processed image data is sent to the diagnostic engine 1450. The image data can be further processed, for example, by the diagnostic engine 1450 and its DDLD 1542 to facilitate the diagnosis of patient 1406….At block 3120, the image / image data is analyzed to determine if the collected image is a good quality image. The data can be compared to one or more thresholds, values, settings, etc. to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient 1406. IQI incorporates a measure of image tolerance for a radiologist (eg, Likert scale) for diagnosis. For example, other indicators such as resolution image quality, noise image quality, biopsy data quality, and / or other data quality metrics can be incorporated to indicate the suitability of the image data for diagnosis. For example, a task-specific data quality index can represent the quality of the collected image data for machine-oriented analysis… if the image / image data quality meets the threshold, the image quality may be evaluated to determine if the quality is too high. Image quality that is too high (eg, IQI of 5 indicating a "complete" image) can indicate that patient 1406 was exposed to too much radiation when collecting image data..--, in [0225]-[0227] of JP 7051849 B2, and in English translated version provided with the Office Action, pages 30-32 {image quality compared to one or more thresholds related to the target clinical application, such as volumes, masses….of tissues, organs…etc.});
LIU (as modified by INAM) and シェイ are combinable as they are in the same field of endeavor: to analysis of medical images {LIU and シェイ both about MR medical images and technologies, such as noises, artificial…etc., in MR image quality} and determination of image quality based on machine learning models. Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify LIU’s method using シェイ’s teachings by including identifying kidney stones, and the another target clinical application comprises identifying gallstones, and wherein the quality threshold associated with the target clinical application is higher than the quality threshold associated with the another target clinical application to LIU’s analysis of the medical image in order to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient (see シェイ: e.g., in English translated version provided with the Office Action, pages 30-32).
Re Claim 8, LIU as modified by INAM disclose that related medical images can be IMR images of “[0026] The medical image represents tissue and/or bone structure of the patient. Alternatively, the medical image represents flow, velocity, or fluids within the patient. In other embodiments, the medical image represents both flow and structure.”,
LIU as modified by INAM however do not explicitly disclose identifying a ventricle, and the another target clinical application comprises identifying pericardial effusion, and wherein the quality threshold associated with the target clinical application is higher than the quality threshold associated with the another target clinical application,
シェイ discloses identifying a ventricle, and the another target clinical application comprises identifying pericardial effusion, and wherein the quality threshold associated with the target clinical application is higher than the quality threshold associated with the another target clinical application (see シェイ: e.g., -- The deployed learning device 2050 is trained to respond to good (or sufficient) quality images and provide suggestions to user 1404 when low (or poor) quality images are obtained. Device 2050 recognizes good image quality and proposes the settings used to collect that high quality image in a particular situation as the default settings for that particular situation….. Settings can be evaluated and configured for multiple imaging modalities such as CT, MICT, SPECT, PET, etc., which follow the same process with different inputs for different outputs. Cardiac imaging, nerve perfusion, lung cancer screening, therapeutic response, etc. can be assisted and improved using the component device 2020. For example, if patient 1406 is scheduled for contrast-enhanced liver examination, patients with normal heart size, size, and liver function will use specific contrast agent settings, but the patient's cardiac function. If is low (eg, reduced), a slower bolus (eg, slower injection rate) and a longer diluted contrast medium are set on the imager 1410 to place the contrast medium at a specific location in the patient.--, in English translated version provided with the Office Action, pages 24-25; -- it is effective to estimate IQ directly from the collected clinical images. A specific example evaluates image quality using a feature-based machine learning or deep learning approach called a learning model. In certain examples, task-based image quality (and / or overall image quality index) can be calculated directly from the image of the actual patient and / or the image of the object.
Images with known image quality (IQ) of interest (eg, clinical images) can be used to train the training model. Additional training images can be generated by manipulating the original image (eg, by blurring or noise insertion to obtain training images with different image qualities). Once the learning model is trained, the model can be applied to new clinical images to estimate the image IQ of interest.
Image input features obtained from cropped raw image data and edge map information, such as mean, standard deviation, kurtosis, skewness, energy, moments, contrast, entropy, etc., are training sets for machine learning systems. Combined with one or more labels such as spatial resolution level, spatial resolution value, etc. to form. The machine learning network uses the training set to form a training model and applies the model to the features obtained from the test set of image data. As a result, the machine learning network outputs an estimated spatial resolution (eg, level and / or value) based on the training model information.--, The data can be compared to one or more thresholds, values, settings, etc. to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient 1406. IQI incorporates a measure of image tolerance for a radiologist (eg, Likert scale) for diagnosis. For example, other indicators such as resolution image quality, noise image quality,
biopsy data quality, and / or other data quality metrics can be incorporated to indicate the suitability of the image data for diagnosis. For example, a task-specific data quality index can represent the quality of the collected image data for machine-oriented analysis-- , and, -- Sources of variation are noise (eg mA, kVp, patient size, etc.), resolution (eg, reconstructed kernel type, thickness, pixel size, etc.), structured noise (eg, streaks, patterns, textures, etc.), shadows. Ing artifacts (eg, ribs, spinal reconstruction artifacts, etc.) can be included. The effect on the result can include measurement error, stage error, and the like. Another use case can include volume / size quantification of other organs (eg, kidney transplants, etc.) or masses within organs (eg, cysts or stones, etc.).--, in [0219] of JP 7051849 B2, and, --Another use case can include cardiac perfusion analysis for diagnosing coronary artery disease (CAD). Sources of variation can include patient physiology (eg, the same patient as the crossing patient, small dynamic range, etc.), beam hardening artifacts (patient uptake, bolus timing, etc.), heart movement, contrast pooling, and the like. Impacts on results can include erroneous perfusion maps (eg, missed perfusion disorders or erroneous diagnosis of perfusion disorders). Another use case can include liver lesions / small dark structures for cancer detection. Sources of variation are noise (eg mA, kVp, patient size, etc.), resolution (eg, reconstructed kernel type, thickness, pixel size, etc.), structured noise (eg, streaks, patterns, textures, etc.), shadows. Ing artifacts (eg, bones, ribs, spinal reconstruction artifacts, etc.), movements, etc. can be included. Impacts on results can include oversight of lesions and misdiagnosis due to poor contrast detectability.
Another use case can include coronary / angiography. Sources of variation can include streaks or blooming artifacts (eg, reconstruction methods, partial volumes, movement, etc.), noise (eg, mA, kVp, patient size, etc.), resolution, and the like. The impact on the results can include the greater impact of noise and resolution when luminal analysis is required.
Another use case can include cerebral perfusion for stroke. Sources of variation can include shadowing artifacts from bone, small physiological changes (eg, small dynamic range, etc.), structured noise (eg, reconstruction methods, etc.), and the like. Impacts on results can include erroneous perfusion maps (eg, missed perfusion disorders or erroneous diagnosis of perfusion disorders).-- ,and, --Similarly, in block 3116, the processed image data is analyzed. For example, image data is analyzed by DDLD1532 for quality, IQI, data quality index, other image quality metrics and the like. The DDLD 1532 learns from the content of machine-processable image data (eg, identified features, resolutions, noise, etc.) and reconstructs previous image data (eg, for the same patient 1406, same type, etc.) and / or reconstruction. Compare with the image. At block 3118, the processed image data is sent to the diagnostic engine 1450. The image data can be further processed, for example, by the diagnostic engine 1450 and its DDLD 1542 to facilitate the diagnosis of patient 1406….At block 3120, the image / image data is analyzed to determine if the collected image is a good quality image. The data can be compared to one or more thresholds, values, settings, etc. to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient 1406. IQI incorporates a measure of image tolerance for a radiologist (eg, Likert scale) for diagnosis. For example, other indicators such as resolution image quality, noise image quality, biopsy data quality, and / or other data quality metrics can be incorporated to indicate the suitability of the image data for diagnosis. For example, a task-specific data quality index can represent the quality of the collected image data for machine-oriented analysis… if the image / image data quality meets the threshold, the image quality may be evaluated to determine if the quality is too high. Image quality that is too high (eg, IQI of 5 indicating a "complete" image) can indicate that patient 1406 was exposed to too much radiation when collecting image data..--, in [0225]-[0227] of JP 7051849 B2, and in English translated version provided with the Office Action, pages 30-33 {image quality compared to one or more thresholds related to the target clinical application, such as volumes, masses….of tissues, organs…etc.});
LIU (as modified by INAM) and シェイ are combinable as they are in the same field of endeavor: to analysis of medical images {LIU and シェイ both about MR medical images and technologies, such as noises, artificial…etc., in MR image quality} and determination of image quality based on machine learning models. Therefore it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify LIU’s method using シェイ’s teachings by including identifying a ventricle, and the another target clinical application comprises identifying pericardial effusion, and wherein the quality threshold associated with the target clinical application is higher than the quality threshold associated with the another target clinical application to LIU’s analysis of the medical image in order to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient (see シェイ: e.g., in English translated version provided with the Office Action, pages 30-32).
Re Claim 9, LIU as modified by INAM and シェイ further disclose wherein when training the machine learning model, the method further comprises: accessing a set of training medical images; labeling each of the set of training medical images with an image quality value; and training the machine learning model using the set of labeled training medical images to predict image quality values for new medical images (see シェイ: e.g., -- The deployed learning device 2050 is trained to respond to good (or sufficient) quality images and provide suggestions to user 1404 when low (or poor) quality images are obtained. Device 2050 recognizes good image quality and proposes the settings used to collect that high quality image in a particular situation as the default settings for that particular situation….. Settings can be evaluated and configured for multiple imaging modalities such as CT, MICT, SPECT, PET, etc., which follow the same process with different inputs for different outputs. Cardiac imaging, nerve perfusion, lung cancer screening, therapeutic response, etc. can be assisted and improved using the component device 2020. For example, if patient 1406 is scheduled for contrast-enhanced liver examination, patients with normal heart size, size, and liver function will use specific contrast agent settings, but the patient's cardiac function. If is low (eg, reduced), a slower bolus (eg, slower injection rate) and a longer diluted contrast medium are set on the imager 1410 to place the contrast medium at a specific location in the patient.--, in English translated version provided with the Office Action, pages 24-25; -- it is effective to estimate IQ directly from the collected clinical images. A specific example evaluates image quality using a feature-based machine learning or deep learning approach called a learning model. In certain examples, task-based image quality (and / or overall image quality index) can be calculated directly from the image of the actual patient and / or the image of the object.
Images with known image quality (IQ) of interest (eg, clinical images) can be used to train the training model. Additional training images can be generated by manipulating the original image (eg, by blurring or noise insertion to obtain training images with different image qualities). Once the learning model is trained, the model can be applied to new clinical images to estimate the image IQ of interest.
Image input features obtained from cropped raw image data and edge map information, such as mean, standard deviation, kurtosis, skewness, energy, moments, contrast, entropy, etc., are training sets for machine learning systems. Combined with one or more labels such as spatial resolution level, spatial resolution value, etc. to form. The machine learning network uses the training set to form a training model and applies the model to the features obtained from the test set of image data. As a result, the machine learning network outputs an estimated spatial resolution (eg, level and / or value) based on the training model information.--, The data can be compared to one or more thresholds, values, settings, etc. to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient 1406. IQI incorporates a measure of image tolerance for a radiologist (eg, Likert scale) for diagnosis. For example, other indicators such as resolution image quality, noise image quality,
biopsy data quality, and / or other data quality metrics can be incorporated to indicate the suitability of the image data for diagnosis. For example, a task-specific data quality index can represent the quality of the collected image data for machine-oriented analysis-- , and, -- Sources of variation are noise (eg mA, kVp, patient size, etc.), resolution (eg, reconstructed kernel type, thickness, pixel size, etc.), structured noise (eg, streaks, patterns, textures, etc.), shadows. Ing artifacts (eg, ribs, spinal reconstruction artifacts, etc.) can be included. The effect on the result can include measurement error, stage error, and the like. Another use case can include volume / size quantification of other organs (eg, kidney transplants, etc.) or masses within organs (eg, cysts or stones, etc.).--, in [0219] of JP 7051849 B2, and, --Another use case can include cardiac perfusion analysis for diagnosing coronary artery disease (CAD). Sources of variation can include patient physiology (eg, the same patient as the crossing patient, small dynamic range, etc.), beam hardening artifacts (patient uptake, bolus timing, etc.), heart movement, contrast pooling, and the like. Impacts on results can include erroneous perfusion maps (eg, missed perfusion disorders or erroneous diagnosis of perfusion disorders). Another use case can include liver lesions / small dark structures for cancer detection. Sources of variation are noise (eg mA, kVp, patient size, etc.), resolution (eg, reconstructed kernel type, thickness, pixel size, etc.), structured noise (eg, streaks, patterns, textures, etc.), shadows. Ing artifacts (eg, bones, ribs, spinal reconstruction artifacts, etc.), movements, etc. can be included. Impacts on results can include oversight of lesions and misdiagnosis due to poor contrast detectability.
Another use case can include coronary / angiography. Sources of variation can include streaks or blooming artifacts (eg, reconstruction methods, partial volumes, movement, etc.), noise (eg, mA, kVp, patient size, etc.), resolution, and the like. The impact on the results can include the greater impact of noise and resolution when luminal analysis is required.
Another use case can include cerebral perfusion for stroke. Sources of variation can include shadowing artifacts from bone, small physiological changes (eg, small dynamic range, etc.), structured noise (eg, reconstruction methods, etc.), and the like. Impacts on results can include erroneous perfusion maps (eg, missed perfusion disorders or erroneous diagnosis of perfusion disorders).-- ,and, --Similarly, in block 3116, the processed image data is analyzed. For example, image data is analyzed by DDLD1532 for quality, IQI, data quality index, other image quality metrics and the like. The DDLD 1532 learns from the content of machine-processable image data (eg, identified features, resolutions, noise, etc.) and reconstructs previous image data (eg, for the same patient 1406, same type, etc.) and / or reconstruction. Compare with the image. At block 3118, the processed image data is sent to the diagnostic engine 1450. The image data can be further processed, for example, by the diagnostic engine 1450 and its DDLD 1542 to facilitate the diagnosis of patient 1406….At block 3120, the image / image data is analyzed to determine if the collected image is a good quality image. The data can be compared to one or more thresholds, values, settings, etc. to determine if the image data collected represents a "good" image. As mentioned above, IQIs, other data quality indices, detectability indices, diagnostic indices, etc. can be generated to represent the reliability and / or usefulness of the data for the diagnosis of patient 1406. IQI incorporates a measure of image tolerance for a radiologist (eg, Likert scale) for diagnosis. For example, other indicators such as resolution image quality, noise image quality, biopsy data quality, and / or other data quality metrics can be incorporated to indicate the suitability of the image data for diagnosis. For example, a task-specific data quality index can represent the quality of the collected image data for machine-oriented analysis… if the image / image data quality meets the threshold, the image quality may be evaluated to determine if the quality is too high. Image quality that is too high (eg, IQI of 5 indicating a "complete" image) can indicate that patient 1406 was exposed to too much radiation when collecting image data..--, in [0225]-[0227] of JP 7051849 B2, and in English translated version provided with the Office Action, pages 30-33). See the similar motivation and obviousness for the combination of cited references statements addressed above for claims 7, and 8.
Re Claims 17-18, claims 17-18 are corresponding system claim to claims 7-8, respectively. Claims 17-18 thus are rejected for the similar reasons for claims 7-8. See above discussions with regard to claims 7-8 respectively. LIU as modified by INAM and シェイ further disclose a system for assessing medical images for suitability in clinical interpretation, the system comprising: a first computing device, at a first location, comprising one or more device processors and a device memory storing device instructions for execution by the one or more device processors, wherein when the device instructions are executed by the one or more device processors, the one or more device processors are configured to perform the method (see LIU: e.g., Fig. 1, -- A scoring system assesses quality and helps determine whether enough significant clinical value may be extracted and therefore lead to correct diagnosis--, in [0002], and, --machine-based deep learning to train a network to predict motion artifacts given an input 3D representation from a scan of a patient. The architecture of the network may be defined to deal with anisotropic data from the MR scan. [0005] In a first aspect, a method is provided for machine learning to determine motion artifacts for a magnetic resonance system. The magnetic resonance system generates a first three-dimensional representation of a patient.--, in [0003]-[0005]; and, --Once the quality score for motion artifact in an MR image volume is predicted, the operator of the medical scanner or the medical scanner decides whether to rescan the patient. The score is used for a decision to use or not use the generated 3D representation. The result is that a later physician review is more likely to have a useful image for diagnosis, and rescanning is avoided where possible. The score may be used to weight an amount of trust in diagnosis based on a 3D representation reconstructed from a scan of the patient.--, in [0073]); and
a second computing device, at a second location, comprising one or more second device processors and a second device memory storing second device instructions for execution by the one or more second device processors, wherein when the second device instructions are executed by the one or more second device processors, the one or more second device processors are configured to perform the method (see Liu: e.g., -- the machine-learned network includes one or more layers with values for various parameters, such as convolution kernels, down sampling weights, and/or connections. The values of the parameters and/or the network as trained are stored in act 19. The machine-learned network is stored in a memory, such as memory of the machine or the database with the examples. The machine-learned network may be transmitted to a different memory. The machine-learned neural network may be duplicated for application by other devices or machines, such as processors of MR scanners. The memories of MR scanners may store copies of the machine-learned network for application for specific patients, enabling a radiologist or other physician to determine whether to rely on an image or to scan again for diagnosis due to patient motion. [0060] FIG. 6 is a flow chart diagram of one embodiment of a method for determining motion artifact for a magnetic resonance system. The stored machine-learned network is applied to determine a score for a scan of a patient. A 3D scan of the patient is performed, and the level of motion artifact from the resulting 3D representation of that patient is determined by the machine-learned network.—in [0059]-[0060]; and, -- The quality score is transmitted over a network, through a communications interface, into memory or database (e.g., to a computerized patient medical record),--, in [0069]-[0070]; and, -- FIG. 7 shows one embodiment of a system for machine learning and/or for application of a machine-learned network. The system is distributed between the imaging system 80 and a remote server 88. In other embodiments, the system is just the server 88--, in [0074]-[0075]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: Accomazzi (US 10839514 B2) discloses a client side viewing application receives the digital images, receives a modification of one of the digital images, and transmits the modified digital image to a server. The server selects a neural network to train based on the digital image, trains the selected neural network, receives a request to apply one of the neural networks to another set of digital images, and selects a neural network to apply.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to WEI WEN YANG whose telephone number is (571)270-5670. The examiner can normally be reached on 8:00 - 5:00 pm.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Amandeep Saini can be reached on 571-272-3382. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/WEI WEN YANG/Primary Examiner, Art Unit 2662
Prosecution Insights