Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Detailed Action
2. This Non-Final Office Action is responsive to Applicants’ reply received 7/10/25. Claims 1-20 were pending, but by way of the reply, claims 5-6 and 15-16 were cancelled and claims 21-22 were added. Hence, claims 1-4, 7-14, and 17-22 are presently pending, of which claims 1 and 11 are independent.
3. The Examiner has found persuasive Applicants’ arguments as relating to previously-pending claim 6 and the Kang reference, the subject matter of which has since been incorporated into the independent claims. Responsive to that, the Examiner has updated the prior art search and is asserting a different reference in place of Kang, which the Examiner reasons is a more fair match with the recited loss function.
4. The Examiner has also withdrawn the previously-made rejection under 35 U.S.C. 101.
Claim Rejections - 35 USC § 103
5. 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.
6. 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.
7. 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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
8. Claims 1-4, 7, 10-14, 17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Esteban et al., “Predicting Clinical Events by Combining Static and Dynamic Information Using Recurrent Neural Networks” in view of Putra et al., “Prediction of Clinical Events in Hemodialysis Patients Using an Artificial Neural Network” and Frank et al., “The Weka Workbench”, and further in view of Xu et al., “Stratified Mortality Prediction of Patients with Acute Kidney Injury in Critical Care” and Grover, “5 Regression Loss Functions All Machine Learners Should Know.”
Regarding Claim 1, Esteban teaches a method for training a predictive model, comprising: training, a dual-channel neural network model which includes a static channel to process static information and a dynamic channel to process temporal information (Esteban, Page 1, Abstract, "In this work we present an approach based on RNNs, specifically designed for the clinical domain, that combines static and dynamic information in order to predict future events"); to generate a probability score that characterizes a likelihood of a health event occurring, based on static profile information and temporal measurement information (Esteban, Page 1, Abstract, “In this work we present an approach based on RNNs, specifically designed for the clinical domain, that combines static and dynamic information in order to predict future events”).
However, while Esteban does teach generating a probability score for a health event as seen above, they fail to teach predicting a health event during a dialysis procedure. Putra does teach generating a probability score of a health event during a dialysis procedure (Putra, Abstract, "We used Artificial Neural Network (ANN) method to predict clinical events during the HD [hemodialysis] sessions"); and training an augmented model to generate an importance score associated with the probability score, based on the static profile information and the temporal measurement information (Putra, Page 1571, Discussion, Paragraph 1, "In our study, MLP, as a feature selection algorithm, extracted HR_FFM as the most importance variable for static feature and age for dynamic feature" also see Table 2 which displays weight values for features). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to predict the likelihood of events defined by Esteban but during the procedure like Putra, and supplement Esteban’s model with a feature ranking model from Putra. The motivation to do so is to support enhanced decision making during a procedure (Putra, Page 1570, Introduction, Paragraph 1, "Despite its benefits, [hemodialysis] comes with risk of muscle spasms, cardiovascular events, and even death [3]" & Paragraph 2, "Artificial neural networks (ANN) have been known to provide state-of-the-art results for most of the classification tasks [4]" & Page 1571, Column 1, Conclusions, Paragraph 1, "Our model could be used to predict the risk of clinical events among HD patients to support decision for healthcare professionals").
The combination of Esteban in view of Petra fail to teach training a neural network using a hardware processor, whereas Frank does (Page 1571, Abstract, "Weka Workbench software was used to train and validate the ANN model" & Frank, Page 89, 4.7 Distributing processing over several machines, "A remarkable feature of the Experimenter is that it can split up an experiment and distribute it across several processors" wherein the Experimenter is the orchestration engine for Weka, and implies that it’s normally tied to at least one processor. Thus, Weka is implemented on at least one processor, and since the model is implemented with Weka, the model is trained on at least one hardware processor). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to train the model proposed by Esteban in view of Putra using a hardware processor. The motivation to do so is to enable the training of a dual channel network model as proposed by Esteban in view of Putra (Putra, Page 1570, Column 1, Abstract, “Weka Workbench software was used to train and validate the ANN model”).
The claim has been amended to include the limitations of cancelled dependent claim 5, e.g. the further limitation wherein training the augmented model includes training gradient boosting trees to output feature importance scores for a static feature from the static channel and a temporal feature from the dynamic channel, which Esteban etc. do not teach. Rather, the Examiner relies upon XU to teach what Esteban etc. otherwise lack, see e.g., Xu: Page 6 The important features chosen from all feature groups, Paragraph 1, "The GBDT model was employed to acquire the importance score of each feature based on 24 hour data" & Page 3, Data pre-processing, Paragraph 1, "We mainly pre-processed two types of features: time-dependent continuous features and discrete features" thus they handled both static and temporal features, and the gradient boosting tree model calculated importance scores for them). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention use Xu’s gradient boosting tree as the augmented model in the combination of Esteban in view of Putra. The motivation to do is to enable appropriate feature weighting (Xu, Page 2, Paragraph 3, "… different features play different roles for mortality prediction in patients with different AKI stages" & Page 6, "Accurate prediction of the mortality risk for AKI patients is helpful for clinicians to understand the condition of the patients and take appropriate actions").
The claim has also been amended to include the limitations of cancelled dependent claim 6, e.g., the further limitation wherein training the augmented model includes by minimizing a loss function for a gradient boosting tree regressor:
l
=
1
N
∑
i
=
1
N
y
^
i
-
(
y
^
n
e
w
)
i
2
2
where N is a number of samples, fj is the probability score for the ith, (ynew)iis a newly obtained probability score for an ithsample. Esteban etc. do not teach this further limitation, and rather the Examiner relies upon GROVER to teach what Esteban etc. otherwise lack, see e.g., Grover’s Mean Square Error / L2 Loss function to consider regression loss on its page 3, under item 1, which the Examiner reasons is the equivalent of Applicants’ recited loss/error function. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use Grover’s loss function as the cost function for training the gradient boosting tree of Esteban in view of Putra and further in view of Xu. The motivation to do so is to enable the benefits of a gradient boosting tree by minimizing the loss/error using a formulation that is well-known and widely-used in the state of the art.
Regarding Claim 11, the claim recites a system comprising a processor and memory with limitations which perform the method of Claim 1, and thus is rejected for the reasons set forth in Claim 1. Frank teaches the usage of a processor and memory (Frank, Page 89, 4.7 Distributing processing over several machines, "A remarkable feature of the Experimenter is that it can split up an experiment and distribute it across several processors" thus using at least one processor for training, wherein the Experimenter is the training orchestration engine & Page 8, 1.2 How do you use it?, Paragraph 2, “… the Explorer … holds everything in main memory” thus the Experimenter uses a memory).
Regarding Claim 2, Esteban in view of Putra and Frank and further in view of Xu and Grover teach the method of claim 1 (and thus the rejection of Claim 1 is incorporated). Esteban further teaches wherein the dynamic channel includes a series of long short-term memory layers (Esteban, Page 2, Column 1, Paragraph 3, "… we developed a new model based on Recurrent Neural Networks" & Page 4, Column 1, Paragraph 5, "Given a sequence of input vectors x = (x1, x2, … xt), the hidden state of an RNN is computed the following way: ht = f(ht-1, xt)." thus, the hidden states of the RNN are time based, and thus is the dynamic channel" & Page 4, Column 2, B. Long Short-Term Memory Units, Paragraph 2, "There are different versions with minor modifications regarding the gating mechanism in the long short-term memory units (LSTM) units. We will use in here the ones defined by Graves et al..." thus they use LSTMs as the hidden states of the RNN for the dynamic channel).
Claim 12 recites a system comprising limitations which perform the method of Claim 2, and thus is rejected for the reasons set forth in Claim 2.
Regarding Claim 3, Esteban in view of Putra and Frank and further in view of Xu and Grover teaches the method of claim 1 (and thus the rejection of Claim 1 is incorporated). Esteban further teaches wherein the static channel includes a multi-layer perceptron (Esteban, Page 5, Column 1, D. Combining RNNs with Static Data, Paragraph 2-3, "The modified architecture is depicted in Figure 3. As it can be seen, we process the static information on an independent Feedforward Neural Network" wherein Feed Forward Neural networks are commonly understood in the field to be multi-layer perceptrons).
Claim 13 recites a system comprising limitations which perform the method of Claim 3, and thus is rejected for the reasons set forth in Claim 3.
Regarding Claim 4, Esteban in view of Putra and Frank and further in view of Xu and Grover teaches the method of claim 1 (and thus the rejection of Claim 1 is incorporated). Esteban further teaches wherein training the dual-channel neural network model includes training a prediction multi-layer perceptron to determine the probability score, based on a static feature from the static channel and a temporal feature from the dynamic channel (Esteban, Figure 3, Page 5, D. Combining RNNs with Static Data, Paragraphs 2-3, "The modified architecture is depicted in Figure 3. As it can be seen, we process the static information on an independent Feedforward Neural Network whereas we process the dynamic information with an RNN. Afterwards we concatenate the hidden states of both networks and provide this information to the output layer" & Column 2, Equations 13-19 which utilize "a vector containing the static information ... a vector containing the information recorded during the visit made by patient i at time t ... we use both hidden states in order to predict our target ... we derive a cost function ...").
Claim 14 recites a system with a computer program product that causes a hardware processor to perform the method of Claim 4, and thus is rejected for the reasons set forth in Claim 4.
Regarding Claim 7, Esteban in view of Putra and Frank and further in view of Xu and Grover teaches the method of claim 1 (and thus the rejection of Claim 1 is incorporated). Esteban further teaches wherein the temporal information includes time series measurements made of one or more characteristics of a patient (Esteban, Page 3, Column 2, Paragraph 3, "Our predictions are based on information from the patient's medical history, the sequence of medications prescribed for the patient, the sequence of laboratory tests performed together with their results").
Claim 17 recites a system comprising limitations which match the method of Claim 7, and thus is rejected for the reasons set forth in Claim 7.
Regarding Claim 10, Esteban in view of Putra and Frank and further in view of Xu and Grover teaches the method of claim 7 (and thus the rejection of Claim 7 is incorporated). Esteban further teaches wherein the temporal information includes blood test information, dialysis measurements, and past event incidences (Esteban, Page 3, Column 2, Paragraph 3, "Our predictions are based on information from the patient's medical history, the sequence of medications prescribed for the patient, the sequence of laboratory tests performed together with their results").
Claim 20 recites a system comprising limitations which match the method of Claim 10, and thus is rejected for the reasons set forth in Claim 10.
9. Claims 8, 9, 18, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Esteban et al., “Predicting Clinical Events by Combining Static and Dynamic Information Using Recurrent Neural Networks” in view of Putra et al., “Prediction of Clinical Events in Hemodialysis Patients Using an Artificial Neural Network” and Frank et al., “The Weka Workbench”, and also Xu and Grover, and yet further in view of Lee, “Predictive Approaches for Acute Adverse Events in Electronic Health Records”.
Regarding Claim 8, Esteban in view of Putra and Frank and further in view of Xu and Grover teaches the method of claim 7 (and thus the rejection of Claim 7 is incorporated). Esteban etc. does not specifically teach, but rather the Examiner relies upon Lee to teach further comprising pre-processing the temporal information, including splitting time series measurements into windows of a predetermined length (Lee, Page 22, 2.2.1. Representation of Patient Physiology from Clinical Data, Paragraph 1, "We represented each patient with multiple instances, in which east instance captured demographic information and clinical information observations within a specific time window"). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to split the temporal information from Esteban in view of Putra in the style performed by Lee. The motivation to do so is to enable enhanced predictive capability (Lee, Page 82, 5.3 Potential usage of the model presented in the dissertation, "It could serve as a baseline modeling approach to quickly evaluate the predictability of the target disease onsets based on the available dataset").
Claim 18 recites a system with a computer program product that causes a hardware processor to perform the method of Claim 8, and thus is rejected for the reasons set forth in Claim 8.
Regarding Claim 9, Esteban in view of Putra and Frank and further in view of Xu and Grover and yet further in view of Lee teaches the method of claim 8 (and thus the rejection of Claim 8 is incorporated). Esteban in view of Putra does not specifically teach, but Lee teaches, wherein pre-processing the temporal information includes adding, as part of a first window, a measurement for a first characteristic that was made outside of the first window, responsive to a determination that the window includes no measurements for the first characteristic (Lee, Page 23, 2.2.1. Representation of Patient Physiology from Clinical Data, Paragraph 3, "When the feature vector could not be populated because of missing observations, we imputed such features with mean values from the training set"). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to replace missing data as Lee does in the temporal dataset of Esteban in view of Putra. The motivation to do so is to enable enhanced predictive capability (Lee, Page 82, 5.3 Potential usage of the model presented in the dissertation, "It could serve as a baseline modeling approach to quickly evaluate the predictability of the target disease onsets based on the available dataset").
Claim 19 recites a system with a computer program product that causes a hardware processor to perform the method of Claim 9, and thus is rejected for the reasons set forth in Claim 9.
Conclusion
10. Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHOURJO DASGUPTA whose telephone number is (571)272-7207. The examiner can normally be reached M-F 8am-5pm CST.
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, Tamara Kyle can be reached at 571 272 4241. 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.
/SHOURJO DASGUPTA/Primary Examiner, Art Unit 2144
Prosecution Insights