Prosecution Insights
Last updated: November 12, 2026
Application No. 17/802,782

TECHNIQUE FOR DETERMINING AN INDICATION OF A MEDICAL CONDITION

Final Rejection §101§103§112
Filed
Aug 26, 2022
Priority
Feb 28, 2020 — EU 20159958.6 +1 more
Examiner
HRANEK, KAREN AMANDA
Art Unit
3684
Tech Center
3600 — Transportation & Electronic Commerce
Assignee
Deepc GmbH
OA Round
5 (Final)
35%
Grant Probability
At Risk
6-7
OA Rounds
73%
With Interview

Examiner Intelligence

Grants only 35% of cases
35%
Career Allowance Rate
70 granted / 198 resolved
-16.6% vs TC avg
Strong +38% interview lift
Without
With
+37.7%
Interview Lift
resolved cases with interview
Typical timeline
Avg Prosecution
24 currently pending
Career history
238
Total Applications
across all art units

Office Action

§101 §103 §112
DETAILED ACTION Notice of Pre-AIA or AIA Status 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 the Claims The status of the claims as of the response filed 7/24/2026 is as follows: Claims 1-15, 23, and 31-32 remain cancelled. Claims 16, 27, 34, and 36 are currently amended. Claims 17-22, 24-26, 28-30, 33, and 35 are as previously presented. Claims 16-22, 24-30, and 33-36 are currently pending in the application and have been considered below. Response to Amendment Rejection Under 35 USC 112(b) Claims 34 and 36 have been sufficiently amended to clarify the indefinite limitations identified in the previous Office action such that the corresponding 35 USC 112(b) rejections are withdrawn. Rejection Under 35 USC 101 The claims have been amended but the 35 USC 101 rejections are upheld. Rejection Under 35 USC 102/103 The amendments made to the claims introduce new limitations that are not fully addressed in the previous Office action, and thus the corresponding 35 USC 102/103 rejections are withdrawn. However, Examiner will consider the amended claims in light of an updated prior art search and address their patentability with respect to prior art below. Response to Arguments Rejection Under 35 USC 101 On page 8 of the response filed 7/24/2026 Applicant argues that determining whether it is possible or impossible to provide a reliable determination of a medical condition based on suitability of at least two selected models as in amended claim 16 “is not a mental observation or subjective judgment” and instead “requires evaluation of whether at least one property associated with the medical data of a test instance indicates suitability of the selected models.” Applicant further asserts that “‘suitability’ is assessed with respect to machine-learning models generated by learning algorithms and configured to generate model-specific indications from medical data” and that “the claim requires execution of the selected models on the medical data to generate model-specific indications,” which a clinician cannot perform in their mind. Applicant concludes that “the execution and gating of trained computational models is materially different from a clinician simply reviewing an image and making a judgment.” Applicant’s arguments are fully considered, but are not persuasive. Examiner first notes that the claims have not been characterized as reciting a mental process, but a certain method of organizing human activity such as managing personal behavior, relationships, and/or interactions between people. Additionally, there are no technical details about how the suitability determinations are being made that would preclude a human actor such as a clinician from managing their personal behavior to perform them. Examiner maintains that a clinician managing their personal behavior would be capable of determining whether a reliable diagnostic output would be possible for at least two models based on suitability of the models for evaluating data with certain properties. For example, a clinician could look at medical data such as lab results, physical examination findings, or images and determine if a reliable determination of a diagnostic output with at least two selected model types is possible or not, e.g. by determining that the lab results include impossible outliers that would skew the results of a diagnostic model, by noting that certain data types needed for the selected models are missing from a physical examination report, by observing that images are of such poor quality that no useful diagnosis can be made with the selected diagnostic models, by confirming that the images centered, cropped, or otherwise formatted as appropriate for the selected models, or some other evaluation of model suitability based on medical data properties. Examiner further notes that the models themselves are only recited in the claims as being “generated by a learning algorithm and configured to provide a model-specific indication of the medical condition based on the medical data.” This description does not inherently limit the models to anything that a human actor could not manage their personal behavior to select and execute; Applicant’s specification even states on Pg4 L11-13 that “the learning algorithm may define a sequence of particular method steps and may be used to optimize the parameters of a statistical model, for example.” Such a disclosure indicates that the models may include simple statistical models as understood by one of ordinary skill in the art, for example a linear regression model that has been trained/fitted to a set of training data via a sequence of particular method steps. Examiner maintains that a clinician managing their personal behavior would be capable of selecting at least two regression equations or other previously fitted/trained simple statistical models, evaluating their suitability with respect to properties of a new test instance, and either notifying a colleague that no reliable diagnosis is possible with the selected models or actually executing the models (e.g. by plugging variables into the regression equation and performing mathematical calculations to determine a diagnostic output) if they determine that a reliable diagnosis is possible. If Applicant intends to limit the models to any specific technical or machine learning architecture, Examiner advises including such limitations in the claims. On page 9 Applicant argues that claim 16 integrates any recited abstract idea into a practical application because it “relates to improving machine-learning-based analysis of medical data.” Applicant asserts that “the operation of the diagnostic system is improved in at least two respects: first, models unsuitable for a specific type of input data are not executed, thereby conserving computational resources; and second, unreliable determinations are not output to the user, thereby improving the quality and accuracy of the diagnostic indication provided.” Applicant’s arguments are fully considered, but are not persuasive. Examiner notes that under the BRI of claim 16, there is no requirement for preventing use of models determined to be unreliable; the only step that occurs if a model is determined to be unreliable is ”triggering output of a notification on an output device, the notification information a user that a reliable determination of the indication is impossible,” which does not actually preclude the unreliable models from being executed in any way. Further, even if some positively recited technical means of preventing execution of unreliable models were required, preventing use of unsuitable diagnostic models is not a technical improvement to a technical field, and instead reflects part of an abstract diagnostic workflow. For example, a clinician might have 10 known diagnostic models or algorithms available to them, and make a determination that 9 of them are not suitable for the specific patient under review based on characteristics of the patient and/or availability of certain types of data, choosing not to execute the 9 unsuitable models and only execute the 1 suitable model, thereby saving time and diagnostic resources and ensuring increased diagnostic accuracy/quality. Just because such an abstract diagnostic and resource conservation process is implemented in a computational environment does not mean that a technical solution to a technical problem is being provided; rather, the abstract resource conservation and diagnostic quality benefits are merely being realized in a computerized application. In other words, because the alleged improvements to efficiency and quality are part of the abstract idea itself, they cannot provide “significantly more” than the abstract idea and thus do not confer eligibility (see MPEP 2106.05(a): “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements.” See also 2106.05(a)(II): “it is important to keep in mind that an improvement in the abstract idea itself… is not an improvement in technology.”). On pages 9-10 Applicant respectfully disagrees that “preventing use of unsuitable diagnostic models is not a technical improvement.” Applicant contends that the claim “addresses a specific technical problem: that different trained models may perform better or worse on different types of medical data, and the difference in performance may not be perceptible to a human user based on the input data alone.” Applicant provides the example of “image noise of different types [that] could degrade a model’s performance, and a user cannot perceive noise in images in a predictable way” such that “the claimed solution involving automatically determining model suitability and gating execution accordingly imposes meaningful limits on the abstract idea and integrates it into a practical application that improves a technological process.” Applicant further analogizes claim 16 to the claims found eligible in Ex parte Desjardins, arguing that “claims reciting specific improvements to how a computer system operates, such as controlling which trained models are executed based on data-driven suitability assessments, are not merely abstract ideas applied on generic hardware.” Applicant’s arguments are fully considered, but are not persuasive. As explained above, Examiner maintains that the alleged improvements to resource usage and diagnostic quality are not technical solutions to technical problems, but instead reflect improvements to abstract diagnostic workflows. Examiner maintains that different trained models performing better or worse on different types of medical data is not inherently a technical problem, and instead reflects the realities of any diagnostic criteria or model; for example, a model capturing the diagnostic criteria for an autism assessment would not be appropriate for evaluating the results of a lipid blood panel, a model capturing the diagnostic criteria for determining whether a patient is experiencing sepsis would not be appropriate for applying to an x-ray image of a broken foot, a model capturing the diagnostic criteria for grading a melanoma would not be appropriate for an MRI of a brain, etc. The problem of selecting diagnostic models appropriate to the properties of input medical data such that a reliable diagnostic indication may be provided is abstract in nature, and the claims as presently drafted do not include technical details about how such a process would be accomplished in a manner that a human actor managing their personal behavior could not achieve (e.g. there is no requirement in claim 16 that the medical data be image data, that the associated be any kind of noise that is undetectable to a human being, that the models are of any specific type that a human actor would not be capable of selecting and executing, etc. as reflected in the unclaimed examples Applicant provides). Examiner further notes that the instant claims are distinct from those found eligible in Desjardins because they do not provide an analogous improvement to how a machine learning model itself is trained and operated. In Desjardins, the claims reflected an improvement to how a machine learning model itself is trained and operates to address the specific technical problem of ‘catastrophic forgetting’ encountered in continual learning systems, which was identified and explained as a technical problem in the specification. In contrast, the instant specification does not outline a specific technical problem in machine learning technology whose solution is reflected in the claims. On page 10 Applicant argues that the Step 2B analysis “does not separately analyze the ordered combination of elements recited in claim 16.” Applicant asserts that the specific ordered combination of elements (1)-(5) “provides a non-conventional technological result – namely, it controls the execution of a diagnostic workflow based on model reliability, saving computational resources that would otherwise be expended executing unreliable models and preventing unreliable diagnoses from being output.” Applicant’s arguments are fully considered, but are not persuasive. As indicated above, many of the features that Applicant points to as providing ‘significantly more’ than the abstract idea itself are actually part of the abstract idea itself, and thus do not provide an inventive concept under Step 2B. Examiner maintains that the only additional elements beyond the abstract idea itself that are recited in claim 16 are (1) that the method is computer-implemented, (2) use of at least one processor to achieve the various steps, and (3) outputting data on an output device. These additional elements, when considered in the context of the claim as a whole, merely serve to automate operations that could occur as a certain method of organizing human activity (as described above), and thus amount to implementation of an abstract idea using generic computer components (see MPEP 2106.05(f)). As evidence of the generic nature of the above recited additional elements, Examiner notes Pg 17 L23 – Pg 18 L7 of Applicant’s specification, where the exemplary computing system is disclosed at a high level in accordance with general-purpose processors, interfaces, and other computing elements known in the art. This disclosure does not indicate that the elements of the invention are particular machines, and instead provides generic examples of computer hardware, such that one of ordinary skill in the art would understand that any generic computer processor and interface / output device could be used to implement the invention. Examiner maintains that the ordered combination of additional elements has been properly evaluated under Step 2B, and that it does not provide an inventive concept or “significantly more” than the judicial exception. On pages 10-11 Applicant argues that amended claim 27 does not recite an abstract idea because it “now requires, inter alia, determining whether an image value exceeds an upper quantile or falls below a lower quantile of the statistical distribution function; identifying a corresponding part of the medical image as anomalous; generating a binary segmentation mask identifying anomalous parts of the medical image; and determining an anomaly score based on a number of anomalous voxels in the medical image” which “constitute image processing having a technical character – specifically, the transformation of medical data into a machine-readable image representation (a binary segmentation mask) of an abnormality” which is “not a process that a clinician could practically perform mentally.” Applicant’s arguments are fully considered, but are not persuasive. Examiner first notes that the steps for identifying a corresponding part of the medical image as anomalous and generating a binary segmentation mask identifying anomalous parts of the medical image are not required to be performed under the BRI of the claim, because such steps would only be performed if the contingent condition of determining that an image value does exceed an upper quantile or fall below a lower quantile of the statistical distribution function is met. In other words, the identifying and mask generation steps would not be required to be performed in a case when no image values are determined as being anomalous. In such a case, the generation of an anomaly score could also include determining that the image is normal or contains no anomalies based on there being no anomalous voxels. Examiner maintains that a human actor managing their personal behavior when performing a diagnostic workflow would be capable of determining whether an image value exceeds an upper quantile or falls below a lower quantile of a statistical distribution, identifying corresponding anomalous parts of the medical image (including identifying no parts when no image values are outside the claimed quantile range are determined), and generating an anomaly score based on the number of anomalous voxel (including determining that the image is normal or has no anomalies based on there being no anomalous voxels identified). Examiner notes that there are no details about what the anomaly score is or how it is determined beyond being broadly “based on a number of anomalous voxels in a medical image” such that a human actor would be capable of observing the number of anomalous voxels in a medical image (e.g. either an absolute or relative quantity) and making some kind of anomaly score determination. Examiner concedes that the contingent step of generating a binary segmentation mask identifying anomalous parts of the image is not part of the abstract idea, but this limitation is addressed as an additional element in Steps 2A – Prong 2 and 2B. On page 11 Applicant argues that “claim 27 represents an improvement in computerized medical image analysis” because “the provision of a binary segmentation mask and quantile-based voxel classification, in the context of the statistical distribution function derived from an atlas-based training image pipeline, improves the field of computer-based medical imaging by enabling statistically robust identification of anomalous image regions.” Applicant’s arguments are fully considered, but are not persuasive. Examiner first notes that the step for generating a binary segmentation mask is not required to occur under the BRI of the claim, such that it does not provide a practical application. Even if it were to be required, Examiner notes that it amounts to a computerized method of highlighting or notifying a user of anomalous regions of a medical image with no technical details about how the mask is generated, such that it amounts to insignificant extra-solution activity because it nominally provides a displayed computerized output of the main analysis step (see MPEP 2106.05(g)). The classification of image values as being anomalous based on comparison to a quantile range is part of the abstract idea itself because a human actor could perform such comparisons by managing their personal behavior (as explained above); as indicated above, an improvement to the abstract idea itself does not provide an improvement to a computer or other technical field and thus does not provide a practical application. On page 11 Applicant argues that in amended claim 27, “the ordered sequence of operations – correlating parts of a medical image with a reference image derived from an atlas, comparing image values to statistical distributions generated from training data, applying quantile-based thresholds to identify anomalous voxels, generating a binary segmentation mask, and determining an anomaly score – defines a specific image processing technique for localizing and classifying abnormalities in volumetric image data” which “constitutes a specific technological implementation, not generic data analysis performed on a conventional computer.” Applicant’s arguments are fully considered, but are not persuasive. As indicated above, many of the features that Applicant points to as providing ‘significantly more’ than the abstract idea itself are actually part of the abstract idea itself, and thus do not provide an inventive concept under Step 2B. Examiner maintains that the only additional elements beyond the abstract idea itself that are recited in claim 27 are (1) that the method is computer-implemented, (2) use of at least one processor to achieve the various steps, and (3) outputting data on an output device. The step of generating a binary segmentation mask is also an additional element, but this step is not required to be performed under the BRI of the claim if the conditions for identifying anomalous parts of the medical image are not met. These additional elements, when considered in the context of the claim as a whole, merely serve to automate operations that could occur as a certain method of organizing human activity (as described above), and thus amount to implementation of an abstract idea using generic computer components (see MPEP 2106.05(f)). As evidence of the generic nature of the above recited additional elements, Examiner notes Pg 17 L23 – Pg 18 L7 of Applicant’s specification, where the exemplary computing system is disclosed at a high level in accordance with general-purpose processors, interfaces, and other computing elements known in the art. This disclosure does not indicate that the elements of the invention are particular machines, and instead provides generic examples of computer hardware, such that one of ordinary skill in the art would understand that any generic computer processor and interface / output device could be used to implement the invention. Further, the generation of a binary segmentation mask as in claim 27 amounts to insignificant extra-solution activity if it is performed, because it merely nominally visually outputs the results of the anomaly detection. Additionally, it is well-understood, routine, and conventional to generate a binary segmentation mask to indicate anomalous portions of a medical image, as evidenced by at least abstract & para. [0076] of Hakuk et al. (US 20170178322 A1); para. [0108] of Davatzikos et al. (US 20160239969 A1); [0047] of Sorensen et al. (US 20060142983 A1); and para. [0107] of Perrin et al. (US 20190156477 A1), such that this feature does not provide an inventive concept or “significantly more” than the abstract idea itself. For the reasons outlined above, the 35 USC 101 rejections are upheld for claims 16-22, 24-30, and 33-36. Rejection Under 35 USC 102/103 On pages 12-13 Applicant alleges various deficiencies of Lee with respect to amended claim 16. Applicant’s arguments are fully considered, but are moot because the new ground of rejection does not rely on Lee for any teaching or matter specifically challenged in the argument. On pages 13-14 Applicant argues that Poole does not teach four aspects of amended claim 27. Examiner agrees that the detection of anomalous voxels via Mahalanobis distance as in Poole is distinct from the quantile-based detection method in steps (1) and (2), but this line of argumentation is moot because the new ground of rejection does not rely upon Poole to teach this subject matter. Applicant’s arguments with respect to limitations (3) and (4) are fully considered, but are not persuasive. Examiner first notes that step (3) is not required to be performed under the broadest reasonable interpretation of the claim because it is contingent upon anomalous parts of the medical image being identified, which would not occur when image values do not exceed an upper quantile or fall below a lower quantile of a statistical distribution function. Regardless, Examiner submits that Poole does sufficiently teach both generating a binary segmentation mask identifying anomalous parts of the medical image and determining an anomaly score based on a number of anomalous voxels in the medical image as presently recited in the claim. Paras. [0021] & [0075] show that locations of detected abnormalities are displayed to a user via highlighting or color-coding of an image, which is considered equivalent to generating a binary segmentation mask identifying anomalous parts of the medical image because each portion of the displayed image is either highlighted or not to show whether it is anomalous or not. Examiner further notes that there is no explanation of what the anomaly score actually is, how it is actually determined, or any other details about the score; para. [0073] of Poole notes that once each voxel has been evaluated to determine if is anomalous, the result is a set of voxels that have been detected as potentially representing abnormalities, with further processing (e.g. connected component analysis, Markov smoothing, or other suitable methods) being performed to result in zero or more candidate regions of abnormality based on the identified voxels, and some regions with a number of anomalous voxels lower than a user-defined threshold may be suppressed. Any of these processing functions occurring after identification of a number of anomalous voxels are considered equivalent to determination of an “anomaly score” that is “based on a number of anomalous voxels in the medical image” because there are no details about what the score is or how it is determined; thus the identification of zero or more candidate regions and/or the suppression of candidate regions with fewer than a defined number of anomalous voxels are considered to encompass “an anomaly score” as presently recited in the claim. On pages 14-15 Applicant argues that Examiner has not provided “an articulated rationale for why a person of ordinary skill in the art would have combined” machine-learning concepts from radiology as in Lee and Zhang with those from multispectral tissue monitoring in surgery as in Maier-Hein, and concludes that “the rejections of claims 34 and 35 rely on impermissible hindsight reconstruction.” Applicant’s arguments are fully considered, but are not persuasive. It must be recognized that any judgment on obviousness is in a sense necessarily a reconstruction based upon hindsight reasoning. But so long as it takes into account only knowledge which was within the level of ordinary skill at the time the claimed invention was made, and does not include knowledge gleaned only from the applicant's disclosure, such a reconstruction is proper. See In re McLaughlin, 443 F.2d 1392, 170 USPQ 209 (CCPA 1971). Examiner maintains that a rationale was provided to explain why a PHOSITA would have found it obvious to combine such features; see at least paras. 57-60 of the non-final rejection mailed 1/27/2026. To summarize, Lee and Zhang show use of various diagnostic models to provide indications of a medical condition (e.g. diagnosis of a lesion) by analyzing medical images, but fail to explicitly disclose that the indication is a score indicating whether an unseen test instance is in- or out-of-distribution compared to a distribution of training data. Paras. [0078]-[0080] & [0355]-[0357] of Maier-Hein teach evaluating medical images with selected image analysis models, including using a model to provide a score indicating whether an unseen test instance is in- or out-of-distribution compared to a distribution of training data. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the medical image analysis methods of the combination to include use of a model providing an out-of-distribution detection score as in Maier-Hein in order to provide information about the closeness of a new image to the set of training data used to train the analysis model so that the algorithm’s accuracy can be confirmed and any spurious results that would not be meaningful can be avoided (as suggested by Maier-Hein [0079] & [0355]-[0357]). The benefit of ensuring or confirming accuracy of a trained model based on understanding the correspondence between a new test instance and the distribution of a training set would be observed when applied to any trained medical analysis model, such that a PHOSITA would have found it obvious to utilize this methodology in the context of radiology image analysis even though it is disclosed in the context of tissue monitoring image analysis. Claim Interpretation Note: claim 27 includes the contingent limitations “identifying a corresponding part of the medical image as anomalous when the image value exceeds the upper quantile or falls below the lower quantile” and “generating a binary segmentation mask identifying anomalous pats of the medical image.” These limitations would only be performed if the contingent condition of the image value exceeding the upper quantile or falling below the lower quantile of the statistical distribution function is met. Per MPEP 2111.04(II): “The broadest reasonable interpretation of a method (or process) claim having contingent limitations requires only those steps that must be performed and does not include steps that are not required to be performed because the condition(s) are not met.” Accordingly, the steps for identifying anomalous image parts and generating a binary segmentation mask identifying the anomalous parts of the medical image are not required under the BRI of claim 27 in a case where it is determined that image values do not exceed the upper quantile or fall below the lower quantile of the statistical distribution function. 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. Claim 36 is 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 applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. Claim 36 recites “the reference image” in line 4. There is insufficient antecedent basis for this limitation in the claim, because parent claim 16 does not introduce a reference image. For purposes of examination, Examiner will interpret “the reference image” as “a reference image” newly introduced by the claim. Claim Rejections - 35 USC § 101 35 U.S.C. 101 reads as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title. Claims 16-22, 24-30, and 33-36 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. Step 1 In the instant case, claims 16-22, 24-27, and 33-36 are directed to methods (i.e. processes), claim 28 is directed to an apparatus (i.e. a machine), and claims 29-30 are directed to a non-transitory computer program product (i.e. a manufacture). Thus, each of the claims falls within one of the four statutory categories. Nevertheless, the claims fall within the judicial exception of an abstract idea. Step 2A – Prong 1 Independent claim 16 recites steps that, under their broadest reasonable interpretations, cover certain methods of organizing human activity, e.g. managing personal behavior, relationships, or interactions between people. Specifically, claim 16 recites: A computer-implemented medical data processing method for determining an indication of a medical condition, the method comprising: selecting, using at least one processor, based on at least one property associated with medical data of a test instance, at least two models out of a plurality of models, wherein each of the plurality of models is generated by a learning algorithm and configured to provide a model-specific indication of the medical condition based on the medical data; determining, using the at least one processor, whether providing a reliable determination of the indication of the medical condition using the selected models is possible or impossible based on suitability of the at least two models, wherein if the at least one property associated with the medical data of the test instance does not indicate suitability of the at least two models such that a reliable determination of the indication is impossible: triggering output of a notification on an output device, the notification informing a user that a reliable determination of the indication is impossible; or wherein, if it is determined that a reliable determination of the indication of the medical condition is possible: determining, using the at least one processor, using each of the selected models, a respective model-specific indication of the medical condition based on the medical data; and determining, using the at least one processor, based on the model-specific indications, the indication of the medical condition. But for the recitation of generic computer components like a computer, a processor, and an output device, the italicized functions, when considered as a whole, describe a model selection and diagnostic process that could take place by a human actor such as a clinician managing their personal behavior and/or interactions with others. For example, a clinician could evaluate medical data in a patient test instance (e.g. a diagnostic image) for a property (e.g. imaging modality) and select appropriate diagnostic models (e.g. those that have previously been fitted or trained using a learning algorithm) based on the property (e.g. choose two or more models that are specific to MRI data when the imaging modality is MRI). The clinician could then look at the images and determine if a reliable determination of a diagnostic output with the selected models is possible or not, e.g. by observing that the images are of such poor quality that no useful diagnosis can be made with the selected diagnostic models, by confirming that the images centered, cropped, or otherwise formatted as appropriate for the selected models, or some other evaluation of model suitability based on image properties. If no reliable diagnosis is possible, the clinician could notify a colleague that diagnosis is impossible; or if a reliable diagnosis is possible, they could use the selected models to determine the indication of a medical condition (e.g. cancerous tumor) based on evaluating the model-specific outputs. Thus, the steps recited in this claim describe instructions that a human actor could follow to manage their personal behavior and/or interactions with others during a diagnostic workflow, and accordingly claim 16 recites an abstract idea in the form of a certain method of organizing human activity. Independent claim 27 also recites steps that, under their broadest reasonable interpretations, cover certain methods of organizing human activity, e.g. managing personal behavior, relationships, or interactions between people. Specifically, claim 27 recites: A computer-implemented medical data processing method for determining an indication of a medical condition, the method comprising: correlating, using at least one processor, parts of a medical image comprised in medical data of a test instance with parts of a reference image; and comparing, using the at least one processor, an image value of at least one part of the medical image with information associated with a correlated part of the reference image to obtain the indication of the medical condition, wherein the information has been generated by: matching, using the at least one processor, a plurality of training images to a base image to correlate parts of each of the training images with parts of the base image, wherein the base image is an atlas image generated based on a further plurality of images; determining, using the at least one processor, image values of at least one part of each of the plurality of training images correlated with a part of the base image, wherein the part of the base image is assigned to the correlated part of the reference image using a predetermined transformation; and determining, using the at least one processor, the information based on the determined image values of the at least one part of each of the plurality of training images, wherein the information is a statistical distribution function of image values of the at least one part of the plurality of training images; wherein obtaining the indication of the medical condition comprises determining whether the image value exceeds an upper quantile or falls below a lower quantile of the statistical distribution function; identifying a corresponding part of the medical image as anomalous when the image value exceeds the upper quantile or falls below the lower quantile; generating a binary segmentation mask identifying anomalous parts of the medical image; determining an anomaly score based on a number of anomalous voxels in the medical image; and triggering output of a notification of the indication on an output device. But for the recitation of generic computer components like a computer, a processor, and an output device, the italicized functions, when considered as a whole, describe a data correlation and determination process that could take place by a human actor such as a clinician managing their personal behavior and/or interactions with others. For example, a clinician could evaluate medical images to correlate test information with reference images from an atlas by comparing image values to statistical distributions in example cases to determine an indication of a medical condition, including comparing image values to upper and lower quantiles of a known statistical distribution to determine if anomalies exist or not (e.g. determining that a mass with certain darkness values in a certain quadrant of an MRI scan indicates a cancerous tumor as indicated by known statistical distributions in of darkness values in reference cases, or that the mass falls within a normal range of darkness values and thus is not anomalous). The clinician could also observe the image to assign an anomaly score based on a number of anomalous voxels they are seeing in the image, e.g. by determining that the image is normal if they do not see anything out of the ordinary. Finally, the clinician could notify a colleague or other person of the diagnostic indication. Thus, the steps recited in this claim describe instructions that a human actor could follow to manage their personal behavior and/or interactions with others, and accordingly claim 27 recites an abstract idea in the form of a certain method of organizing human activity. Dependent claims 17-22, 24-26, and 28-30, and 33-36 inherit the limitations that recite an abstract idea from their dependence on claim 16, and thus these claims also recite an abstract idea under the Step 2A – Prong 1 analysis. In addition, claims 17-22, 24-26, and 33-36 recite additional limitations that further describe the abstract idea identified in the independent claim. Specifically, claims 17-18 specify the type of property used as a basis of selection for the models, each of which are types of properties that a clinician would be capable of evaluating as a basis for selecting diagnostic models. Claim 19 recites that the step of selecting the models includes comparing the property with a property associated with training data for each individual model, which a clinician could accomplish by choosing models based on the types of data used to initialize/train them (e.g. selecting a model specialized for women over the age of 50 when a patient has those characteristics). Claim 20 recites that the indication is determined based on at least one attribute chosen from a result of the comparing, performances of each of the models, and a degree of explainability of each model. A clinician could determine the medical indication based on the comparison or a performance of each model by selecting and using a model that most closely matches the desired type of input data or has the best performance metrics. Claim 21 recites that at least one of the models is optionally configured to provide an anomaly detection as the indication, which is a type of model that a human actor would be capable of selecting and using. Claim 22 describes several types of indications that may be output by the model, each of which are types of outputs that a clinician would be capable of gleaning from a selected diagnostic model. Claims 24-26 recite correlating parts of a medical image with parts of a referencing image, and comparing an image value of part of the image with previously generated information associated with a correlated part of the reference image (e.g. statistical distribution function or average image value) to obtain the model-specific indication of the medical condition. These limitations describe a data correlation and determination process that could take place by a human actor such as a clinician managing their personal behavior and/or interactions with others. For example, a clinician could evaluate medical images to correlate test information with reference images from an atlas by comparing image values to statistical distributions in example cases to determine an indication of a medical condition (e.g. determining that a mass with certain darkness values in a certain quadrant of an MRI scan indicates a cancerous tumor as indicated by known statistical distributions or average values in of darkness values in reference cases). Claim 33 specifies that the indication of the medical condition is a medical diagnosis, which a clinician would be capable of gleaning from analysis of a medical image and/or other patient data. Claims 34-35 specify that outputs of a model include out-of-distribution detection metrics, which a clinician could calculate by comparing the test instance to a set of data used to fit each model and determining whether the test instance is similar to the training data or an outlier (e.g. by noting that a given diagnostic model was fitted using data from patients aged 50-70, but a current test instance relates to data for a patient who is 4 years old and is thus out-of-distribution for that model). Claim 36 recites that at least one of the models is a density model configured to compare an image value of the medical image with a correlated reference image to obtain a model-specific indication of a medical condition, which a clinician could achieve by utilizing density-related image values when comparing a new image to a reference image to output a diagnostic indication. However, recitation of an abstract idea is not the end of the analysis. Each of the claims must be analyzed for additional elements that indicate the abstract idea is integrated into a practical application to determine whether the claim is considered to be “directed to” an abstract idea. Step 2A – Prong 2 The judicial exception is not integrated into a practical application. In particular, independent claims 16 and 27 do not include additional elements that integrate the abstract idea into a practical application. The additional elements of claims 16 and 27 include that each method is computer-implemented, use of at least one processor to achieve the various steps, and outputting data on an output device. These additional elements, when considered in the context of each claim as a whole, merely serve to automate operations that could occur as a certain method of organizing human activity (as described above), and thus amount to implementation of an abstract idea using generic computer components (see MPEP 2106.05(f)). For example, a clinician would be capable of making model selections, comparing and correlating data, and making medical determinations based on test data. Use of a computer or processor to achieve these functions then amounts to the words “apply it” such that the otherwise-abstract steps are merely digitized and/or automated using generic computing components and do not provide integration into a practical application. Similarly, a clinician would be capable of outputting a notification to a colleague or other human actor as indicated above, and use of an output device as the medium for outputting information amounts to instructions to “apply” the otherwise-abstract step of sharing data between entities in a digital environment. Claim 27 also includes the additional element of generating a binary segmentation mask identifying anomalous parts of the medical image. Examiner first notes that this limitation is not required to be performed under the broadest reasonable interpretation of the claim (see para. 19 above), so it does not confer eligibility. Further, even if it were required to be performed, it amounts to a computerized method of highlighting or notifying a user of anomalous regions of a medical image with no technical details about how the mask is generated, such that it amounts to insignificant extra-solution activity because it nominally provides a displayed computerized output of the main analysis step (see MPEP 2106.05(g)). The judicial exception recited in dependent claims 17-22, 24-26, 28-30, and 33-36 is also not integrated into a practical application under a similar analysis as above. The functions of claims 17-20, 22, 24-26, and 33-36 are performed with the same additional elements introduced in the independent claim, without introducing any new additional elements of their own, and accordingly also amount to mere instructions to apply the abstract idea using these same additional elements. Claim 21 specifies that at least one of the models is generated by an unsupervised learning algorithm using unlabeled training data of health patients, which merely indicates that a high-level type of machine learning algorithm is utilized for at least one of the models to train it prior to the method. Such a high-level indication of machine learning merely services to digitize and/or automate the otherwise-abstract diagnostic model such that it also amounts to the words “apply it” with a computer and does not provide integration into a practical application. Claims 28-30 recite additional high-level computer components such as at least one processor and at least one memory, a non-transitory computer program product, and one or more computer readable recording media for implementing the steps of claim 16, which also amounts to the words “apply it” with a computer as explained for similar high-level computer components recited in claim 16. Accordingly, the additional elements of claims 16-22, 24-30, and 33-36 do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. Claims 16-22, 24-30, and 33-36 are directed to an abstract idea. Step 2B The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of a computer, at least one processor, and other high-level computing components like a memory and output device for performing the selecting, determining, triggering, correlating, comparing, identifying, generating, etc. steps of the invention amount to mere instructions to apply the exception using generic computer components. As evidence of the generic nature of the above recited additional elements, Examiner notes Pg 17 L23 – Pg 18 L7 of Applicant’s specification, where the exemplary computing system is disclosed at a high level in accordance with general-purpose processors, interfaces, and other computing elements known in the art. This disclosure does not indicate that the elements of the invention are particular machines, and instead provide generic examples of computer hardware, such that one of ordinary skill in the art would understand that any generic computer processor, memory, and interface could be used to implement the invention. The generation of a binary segmentation mask as in claim 27 amounts to insignificant extra-solution activity because it merely nominally visually outputs the results of the anomaly detection, as indicated above. Additionally, it is well-understood, routine, and conventional to generate a binary segmentation mask to indicate anomalous portions of a medical image, as evidenced by at least abstract & para. [0076] of Hakuk et al. (US 20170178322 A1); para. [0108] of Davatzikos et al. (US 20160239969 A1); [0047] of Sorensen et al. (US 20060142983 A1); and para. [0107] of Perrin et al. (US 20190156477 A1). Further, the use of unsupervised machine learning as in claim 21 does not appear to be an inventive concept, as evidenced by the high-level explanation of known unsupervised learning techniques on Pg 9 L21 – Pg 10 L10 of Applicant’s specification. Examiner also notes that it is well-understood, routine, and conventional to utilize unsupervised machine learning for the purpose of medical diagnosis and anomaly detection, as evidenced by at least para. [0116] of Nachev (US 20130304710 A1); para. [0032] of Nye et al. (US 20190150857 A1); and para. [0051] of Zhang et al. (US 20210042916 A1). Analyzing these additional elements as an ordered combination adds nothing that is not already present when considering the elements individually; the overall effect of the computer implementation, unsupervised learning model, and output device in combination is to digitize and/or automate medical model selection and comparative diagnostic determination operations that could otherwise be achieved as a certain method of organizing human activity. Thus, when considered as a whole and in combination, claims 16-22, 24-30, and 33-36 are not patent eligible. 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 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 16-20, 22, 28-30, and 33 are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al. (US 20140101080 A1) in view of Zhang et al. (Reference U on the accompanying PTO-892, hereinafter “Zhang1”). Claim 16 Lee teaches a computer-implemented medical data processing method for determining an indication of a medical condition (Lee abstract, [0112], noting computerized methods of diagnosing a lesion in a medical image, i.e. determining an indication of a medical condition), the method comprising: selecting, using at least one processor, based on at least one property associated with medical data of a test instance, at least two models out of a plurality of models, wherein each of the plurality of models is generated by a learning algorithm and configured to provide a model-specific indication of the medical condition based on the medical data (Lee Fig. 9, [0055], noting model selection unit searches for and automatically selects one or more (e.g. at least two as shown in the output of Fig. 7) categorized diagnostic models suitable for diagnosing a new patient based on extracted patient information and/or medical image data (i.e. at least one property associated with medical data of a test instance such as a diagnostic image) of the new patient; each diagnostic model is generated via machine learning and configured to provide a model-specific indication of a lesion as noted in [0053] & [0056]); (Lee Figs. 7 & 9, [0056], noting each selected suitable model performs a diagnosis of a lesion in a patient image); and determining, using the at least one processor, based on the model-specific indications, the indication of the medical condition (Lee Figs. 7 & 9, [0057], noting the system determines an integrated diagnosis result based on the model-specific indications). In summary, Lee teaches a method for selecting and executing diagnostic models suitable to evaluate medical image data based on properties of the medical images. Though it contemplates execution of the selected suitable models, it fails to explicitly disclose a distinct prior step for determining whether a reliable determination of the indication of the medical condition using the selected models is possible or impossible based on suitability of the at least two models, and either executing the models specifically if it is determined that a reliable determination of the indication of the medical condition is possible, or triggering output of a notification informing a user that a reliable determination of the indication is impossible on an output device if the at least one property associated with the medical data of the test instance does not indicate suitability of the at least two models such that a reliable determination of the indication is impossible. However, Zhang1 teaches a model-agnostic method of determining whether a reliable determination of a desired computer vision task is possible based on suitability of various image analysis models to evaluate input images based on their various properties (Zhang1 Pgs1-3). The reliability determination step precedes execution of the evaluated models, and can include outputting a warning to a user when a reliable determination is impossible (Zhang1 end of Pg1 Col2 – beginning of Pg2 Col1) as well as later execution of the models to perform the desired computer vision task, which would include cases when a reliable determination is possible (Zhang1 Pgs1-2). It therefore would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the suitable model selection and execution method of Lee to include a prior step for determining whether a reliable determination is possible with the models resulting in either a warning or execution of the models as in Zhang1 in order to make the overall image analysis pipeline more robust and accurate by allowing the models to make fewer, but more accurate decisions, as well as to allow a user to make an informed decision about whether to trust the output of the executed model(s) (as suggested by Zhang1 end of Pg1 Col2 – beginning of Pg2 Col1 & Pg2 end of Col2). The result of such a combination would include the selection of suitable diagnostic models based on image properties as in Lee, followed by the evaluation of whether a reliable output is possible for each suitable model for the image properties of a given input image instance as in Zhang1, followed by either the output of a warning to a user that a reliable determination is impossible as in Zhang1 or the execution of models for which a reliable output is possible as in Lee & Zhang1. Claim 17 Lee in view of Zhang1 teaches the method of claim 16, and the combination further teaches wherein the at least one property associated with the medical data comprises a feature of a medical image comprised in the medical data (Lee [0052], [0055], noting the categories of data used as a basis for training different models and selecting a suitable model for a patient include a feature of a medical image such as imaging device information). Claim 18 Lee in view of Zhang1 teaches the method of claim 16, and the combination further teaches wherein the at least one property associated with the medical data comprises a characteristic of a patient to which the medical data relates (Lee [0052], [0055], noting the categories of data used as a basis for training different models and selecting a suitable model for a patient include a feature of patient data such as clinical information, demographics, etc.). Claim 19 Lee in view of Zhang1 teaches the method of claim 16, and the combination further teaches wherein the step of selecting comprises comparing, using the at least one processor, the at least one property associated with the medical data of the test instance with at least one property associated with training data used for generating an individual model, for each individual model of the plurality of models (Lee Fig. 9, [0055], noting model selection unit searches for and automatically selects the categorized diagnostic models suitable for diagnosing a new patient based on matching (i.e. comparing) extracted patient information and/or medical image data (i.e. at least one property associated with medical data of a test instance such as a diagnostic image) with categories of data used to train each category-specific model (e.g. as in [0053])). Claim 20 Lee in view of Zhang1 teaches the method of claim 19, and the combination further teaches wherein the indication is determined further, using the at least one processor, based on at least one attribute chosen from a result of the comparing, empirical performances of each of the plurality of models and a degree of explainability of each of the plurality of models (Lee [0057], noting the integrated diagnosis result is determined at least based on the results of each selected categorized model (i.e. based on a result of the comparing step of claim 19 indicating that such a model should be selected for diagnosis), and/or based on preference data for each model that may indicate high or low precision of a model (i.e. empirical performances of each of the models)). Claim 22 Lee in view of Zhang1 teaches the method of claim 16, and the combination further teaches wherein the model-specific indication of the medical condition and/or the indication of the medical condition comprises at least one result chosen from probabilities of an anomaly for different parts of a medical image comprised in the medical data and a numerical value describing a probability of an anomaly of the overall medical data, wherein the numerical value is optionally derived from the probabilities of the anomaly for the different parts of the medical image (Lee Fig. 7, [0057], noting the integrated diagnosis result (i.e. indication of the medical condition) includes a numerical probability of the lesion in an image being benign or malignant, i.e. an anomaly). Claim 28 Lee in view of Zhang1 teaches an apparatus comprising at least one processor and at least one memory, the at least one memory containing instructions executable by the at least one processor such that the apparatus unit is operable to perform the method of claim 16 (Lee [0112]-[0113]). Claim 29 Lee in view of Zhang1 teaches a non-transitory computer program product comprising program code portions for performing the method of claim 16 when the computer program product is executed on one or more processors (Lee [0112]-[0113]). Claim 30 Lee in view of Zhang1 teaches the non-transitory computer program product of claim 29, stored on one or more computer readable recording media (Lee [0112]-[0113]). Claim 33 Lee in view of Zhang1 teaches the computer-implemented medical data processing method of claim 16, and the combination further teaches wherein the indication of the medical condition is a medical diagnosis (Lee [0056], noting the models may diagnose a lesion of a patient as malignant or benign). Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Lee and Zhang1 as applied to claim 16 above, and further in view of Zhang et al. (US 20210042916 A1, hereinafter “Zhang2”). Claim 21 Note: claim 21 includes the optional limitation “optionally, configured to provide an anomaly detection as the model-specific indication of the medical condition.” Per MPEP 2111.04(I): “Claim scope is note limited by claim language that suggests or makes optional but does not require steps to be performed, or by claim language that does not limit a claim to a particular structure.” Accordingly, the optional limitation cited above need not be taught by the prior art under the broadest reasonable interpretation of claim 21. However, in the interest of compact prosecution, this limitation is addressed with prior art below. Lee in view of Zhang1 teaches the method of claim 16, and the combination further teaches wherein at least one of the models comprised in the plurality of models is generated, using the at least one processor, by an (Lee [0066], [0092], noting a variety of learning algorithms may be used to generate the categorized models using training data of both malignant (i.e. diseased) and benign (i.e. healthy) patients; the models are trained to distinguish between malignant and benign lesions, which is considered equivalent to providing anomaly detection because malignant lesions are an anomalous medical indication). In summary, the present combination teaches that “nearly all types of machine learning algorithms” may be utilized to train the diagnostic models (see Lee [0066]), but fails to explicitly disclose an unsupervised learning algorithm using unlabeled training data. However, Zhang2 teaches use of an unsupervised learning algorithm using unlabeled training data to train diagnostic models (Zhang2 [0051]). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the learning methods of the combination to include unsupervised learning via unlabeled training data as in Zhang2 because Lee already contemplates “nearly all types of machine learning algorithms” and Zhang2 notes that “unsupervised learning is useful for identifying the features that are most useful for classifying raw data into separate cohorts” (see [0051]). Claim 34-35 are rejected under 35 U.S.C. 103 as being unpatentable over Lee, Zhang1, and Zhang2 as applied to claim 21 above, and further in view of Maier-Hein et al. (US 20220008157 A1). Claim 34 Lee in view of Zhang1 and Zhang2 teaches the computer-implemented medical data processing method of claim 21, and the combination further teaches wherein at least one model is configured to provide the model-specific indication of a medical condition, (Lee Figs. 7 & 9, [0056], noting each selected model performs a diagnosis of a lesion in a patient image). However, the present combination fails to explicitly disclose wherein the model-specific indication is a score indicating whether an unseen test instance is in- or out-of-distribution compared to a distribution of the training data. However, Maier-Hein teaches a method for evaluating medical images with selected image analysis models that includes using a model to provide a score indicating whether an unseen test instance is in- or out-of-distribution compared to a distribution of the training data (Maier-Hein [0078]-[0080], [0355]-[0357]). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the image analysis methods of the combination to include use of a model providing an out-of-distribution detection score as in Maier-Hein in order to provide information about the closeness of a new image to the set of training data used to train an analysis model so that an algorithm’s accuracy can be confirmed and any spurious results that would not be meaningful can be avoided (as suggested by Maier-Hein [0079] & [0355]-[0357]). Claim 35 Note: claim 35 further limits the optional limitation from parent claim 21 regarding providing an anomaly detection as the model-specific indication of a medical condition. Per MPEP 2111.04(I): “Claim scope is note limited by claim language that suggests or makes optional but does not require steps to be performed, or by claim language that does not limit a claim to a particular structure.” Accordingly, all of claim 35 need not be taught by the prior art under the broadest reasonable interpretation of the claim. However, in the interest of compact prosecution, this claim is addressed with prior art below. Lee in view of Zhang1 and Zhang2 teaches the computer-implemented medical data processing method of claim 21, but the combination fails to explicitly disclose wherein the anomaly detection corresponds to an out-of-distribution detection. However, Maier-Hein teaches a method for evaluating medical images with selected image analysis models that includes using a model to provide anomaly detection such as out-of-distribution detection (Maier-Hein [0078]-[0080], [0355]-[0357]). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the image analysis methods of the combination to include use of a model providing out-of-distribution detection as in Maier-Hein in order to provide information about the closeness of a new image to the set of training data used to train an analysis model so that an algorithm’s accuracy can be confirmed and any spurious results that would not be meaningful can be avoided (as suggested by Maier-Hein [0079] & [0355]-[0357]). Claims 24-26 and 36 are rejected under 35 U.S.C. 103 as being unpatentable over Lee and Zhang1 as applied to claim 16 above, and further in view of Poole (US 20130044927 A1). Claim 24 Lee in view of Zhang1 teaches the method of claim 16, and the combination further shows use of quantitative feature values extracted from images to obtain the model-specific indications (Lee [0092]). However, the present combination does not teach correlation to reference images and comparison of the extracted quantitative feature values to information from the reference image, and thus fails to explicitly disclose: wherein at least one of the models comprised in the plurality of models correlates parts of a medical image comprised in the medical data with parts of a reference image, and compares an image value of at least one part of the medical image with information associated with a correlated part of the reference image to obtain the model-specific indication of the medical condition, wherein the information has been generated by: matching a plurality of training images to a base image to correlate parts of each of the training images with parts of the base image; determining image values of at least one part of each of the plurality of training images correlated with a part of the base image, wherein the part of the base image is assigned to the correlated part of the reference image using a predetermined transformation; and determining the information based on the determined image values of the at least one part of each of the plurality of training images. However, Poole teaches an analogous medical image diagnostic method in which: a diagnostic model correlates parts of a medical image with parts of a reference image and compares an image value of at least one part of the medical image with information associated with a correlated part of the reference image to obtain an indication of a medical condition (Poole [0021], [0062]-[0069], noting a patient image data set is registered (i.e. correlated) to a statistical atlas and values of each voxel (i.e. part) of the registered patient image and atlas image are compared to detect abnormalities (i.e. an indication of a medical condition)), wherein the information has been generated by: matching a plurality of training images to a base image to correlate parts of each of the training images with parts of the base image (Poole [0038]-[0047], noting training images are iteratively aligned with a statistical atlas (i.e. base image) that has been generated from previous rounds of iterative alignment); determining image values of at least one part of each of the plurality of training images correlated with a part of the base image (Poole [0029]-[0039], noting the method maintains a mean vector and covariance matrix for each voxel of the registered training images representing texture values of each aligned voxel in the correlated training images), wherein the part of the base image is assigned to the correlated part of the reference image using a predetermined transformation (Poole [0038]-[0039], noting the training images are registered (i.e. correlated) to the current iteration of the statistical atlas (i.e. the base image) using a predetermined transformation Tk); and determining the information based on the determined image values of the at least one part of each of the plurality of training images (Poole [0025], [0037], [0058], noting the method obtains mean and inverted covariance matrices representing statistical distributions (e.g. multivariate Gaussian models) of texture values of each aligned voxel in the correlated training images). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the diagnostic image analysis method of Lee to include a specific method based on alignment and comparison with a generated atlas as in Poole in order to utilize widely practiced diagnosis methods that enable direct comparisons to be performed between image data obtained from different subjects so that more efficient identification and highlighting of abnormal regions of interest may be facilitated (as suggested by Poole [0002] & [0004]). Claim 25 Lee in view of Zhang1 and Poole teaches the method of claim 24, and the combination further teaches wherein the information comprises or is a statistical distribution function of image values of the at least one part of the plurality of training images (Poole [0025], [0037], [0058], noting the atlas data includes mean and inverted covariance matrices representing statistical distributions (e.g. multivariate Gaussian models) of texture values of each aligned voxel in the correlated training images). Claim 26 Lee in view of Zhang1 and Poole teaches the method of claim 24, and the combination further teaches wherein the information comprises an average image value of the at least one part of all of the plurality of training images and, optionally, a mean deviation of the image values of the at least one part of all of the plurality of training images from the average image value (Poole [0025], [0037], [0058], noting the atlas data includes mean (i.e. average) and inverted covariance matrices representing texture values of each aligned voxel in the correlated training images). Claim 36 Lee in view of Zhang1 teaches the computer-implemented medical data processing method of claim 16, but the combination fails to explicitly disclose wherein the at least one model in the plurality of models is a density model configured to compare an image value of at least one part of the medical image comprised in or represented by the medical data with information associated with a correlated part of the reference image to obtain the model-specific indication of the medical condition. However, Poole teaches an analogous medical image diagnostic method in which at least one model in the plurality of models is a density model configured to compare an image value of at least one part of the medical image comprised in or represented by the medical data with information associated with a correlated part of the reference image to obtain the model-specific indication of the medical condition (Poole [0021], [0064]-[0069], [0073], noting abnormality detection module compares values each voxel of the registered patient image and atlas image to detect abnormalities (i.e. indications of a medical condition); the abnormality detection module is thus considered equivalent to a density model as disclosed on Pg 12 of Applicant’s specification and having the configuration recited in this claim because it performs equivalent functions). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the diagnostic image analysis models of the combination to include a model that compares an image value with a correlated part of a reference image to perform the diagnostic output as in Poole in order to utilize widely practiced diagnosis methods that enable direct comparisons to be performed between image data obtained from different subjects so that more efficient identification and highlighting of abnormal regions of interest may be facilitated (as suggested by Poole [0002] & [0004]). Claim 27 is rejected under 35 U.S.C. 103 as being unpatentable over Poole in view of Chen et al. (US 20190318471 A1). Claim 27 Poole teaches a computer-implemented medical data processing method for determining an indication of a medical condition (Poole abstract, [0015]-[0016], noting a processor-based computer system for detecting the presence of abnormalities in medical images, i.e. determining an indication of a medical condition), the method comprising: correlating, using at least one processor, parts of a medical image comprised in medical data of a test instance with parts of a reference image (Poole [0021], [0062]-[0063], noting a patient image data set is registered to a statistical atlas, i.e. parts of a medical image are correlated with parts of a reference image); and comparing, using the at least one processor, an image value of at least one part of the medical image with information associated with a correlated part of the reference image to obtain the indication of the medical condition (Poole [0021], [0064]-[0069], [0073], noting values of each voxel of the registered patient image and atlas image are compared to detect abnormalities, i.e. an image value of at least one part of the medical image is compared with information associated with a correlated part of the reference image to obtain an indication of a medical condition), wherein the information has been generated by: matching, using the at least one processor, a plurality of training images to a base image to correlate parts of each of the training images with parts of the base image, wherein the base image is an atlas image generated based on a further plurality of images (Poole [0038]-[0047], noting training images are iteratively aligned with a statistical atlas (i.e. base image) that has been generated from previous rounds of iterative alignment); determining, using the at least one processor, image values of at least one part of each of the plurality of training images correlated with a part of the base image (Poole [0029]-[0039], noting the method maintains a mean vector and covariance matrix for each voxel of the registered training images representing texture values of each aligned voxel in the correlated training images), wherein the part of the base image is assigned to the correlated part of the reference image using a predetermined transformation (Poole [0038]-[0039], noting the training images are registered (i.e. correlated) to the current iteration of the statistical atlas (i.e. the base image) using a predetermined transformation Tk); determining, using the at least one processor, the information based on the determined image values of the at least one part of each of the plurality of training images, wherein the information is a statistical distribution function of image values of the at least one part of the plurality of training images (Poole [0025], [0037], [0058], noting the method obtains mean and inverted covariance matrices representing statistical distributions (e.g. multivariate Gaussian models) of texture values of each aligned voxel in the correlated training images); wherein obtaining the indication of the medical condition comprises determining whether the image value exceeds an outlier metric (Poole [0064]-[0070], noting a value of each voxel is compared to a statistical distribution of the reference images to determine if it is an outlier); identifying a corresponding part of the medical image as anomalous when the image value exceeds the outlier metric (Poole [0064]-[0070], noting a value of each voxel is compared to a statistical distribution of the reference images to determine if it is an outlier and therefore is identified as being anomalous); generating a binary segmentation mask identifying anomalous parts of the medical image (Poole [0021], [0075], noting locations of detected abnormalities are displayed to a user (e.g. on the display device of [0015]) via highlighting or color-coding, considered equivalent to generation of a binary segmentation mask identifying anomalous parts of the medical image); determining an anomaly score based on a number of anomalous voxels in the medical image (Poole [0073], noting once each voxel has been evaluated to determine if is anomalous, the result is a set of voxels that have been detected as potentially representing abnormalities; further processing (e.g. connected component analysis, Markov smoothing, or other suitable methods) is performed to result in zero or more candidate regions of abnormality based on the identified voxels, and some regions with a number of anomalous voxels lower than a user-defined threshold may be suppressed. Any of these processing functions occurring after identification of a number of anomalous voxels are considered equivalent to determination of an “anomaly score” that is “based on a number of anomalous voxels in the medical image” because there is no explanation of what the score actually is, how it is actually determined, or any other details; thus the identification of zero or more candidate regions and/or the suppression of candidate regions with fewer than a defined number of anomalous voxels are considered to encompass “an anomaly score” as presently recited in the claim); and triggering output of a notification of the indication on an output device (Poole [0021], [0075], noting locations of detected abnormalities are displayed to a user (e.g. on the display device of [0015]) via highlighting or color-coding, considered equivalent to triggering output of a notification of the indication to an output device because the user is notified of the detected abnormalities at a display device). In summary, Poole teaches a method for training and using a diagnostic image analysis model to identify anomalous voxels in an image based on comparison to feature distributions in correlated reference images. Though Poole discusses identifying anomalous voxels of an image via statistical outlier detection (e.g. via calculation of Mahalanobis distance with reference to a distribution of values in a reference image as in [0064]-[0069]), it fails to explicitly disclose detecting the outliers specifically when an image value exceeds an upper quantile or falls below a lower quantile of the statistical distribution. However, Chen teaches an analogous image-based anomaly/defect detection method that identifies the anomalous/defective regions of an image by comparing a test image to statistical distributions of image features of a reference image and determining pixel-specific outliers above a third quartile or below a first quartile as the anomalous/defective regions (Chen [0033]-[0034]), showing that this method of image-based outlier detection was known in the art at the time of filing. Since each individual element and its function are shown in the prior art, albeit shown in separate references, the difference between the claimed subject matter and the prior art rests not on any individual element or function but in the very combination itself – that is in the substitution of the quantile-based outlier/anomaly detection method of Chen for the Mahalanobis distance-based outlier/anomaly detection method of Poole. Thus, the simple substitution of one known element for another producing a predictable result (i.e. the statistically-based detection of outliers indicative of anomalous pixels in test images as compared with reference images) renders the claim obvious. Conclusion The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Mishra (US 9465994 B1) describes a system for predicting how well-suited different machine learning algorithms are for evaluating certain input datasets based on attributes of the datasets, such that only those algorithms predicted to have a high level of success or other performance metrics are actually executed. Ikemoto et al. (US 20150181185 A1) describes a system for calculating a lesion index of an image based on determining the number of pixels representing a lesion in the image. 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 KAREN A HRANEK whose telephone number is (571)272-1679. The examiner can normally be reached M-F 8:00-4:00 ET. 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, Shahid Merchant can be reached on 571-270-1360. 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. /KAREN A HRANEK/ Primary Examiner, Art Unit 3684
Read full office action

Prosecution Timeline

Show 3 earlier events
Feb 21, 2025
Non-Final Rejection mailed — §101, §103, §112
May 21, 2025
Response Filed
Jun 13, 2025
Final Rejection mailed — §101, §103, §112
Dec 10, 2025
Request for Continued Examination
Dec 17, 2025
Response after Non-Final Action
Jan 27, 2026
Non-Final Rejection mailed — §101, §103, §112
Jul 24, 2026
Response Filed
Sep 28, 2026
Final Rejection mailed — §101, §103, §112 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12755839
Machine Learning Model Based Condition and Property Detection
4y 4m to grant Granted Oct 06, 2026
Patent 12731689
MEDICAL DIAGNOSIS GENERATION
2y 9m to grant Granted Sep 08, 2026
Patent 12646620
JOINT STATE ESTIMATION PREDICTION THAT EVALUATES DIFFERENCES IN PREDICTED VS. CORRESPONDING RECEIVED DATA
5y 6m to grant Granted Jun 02, 2026
Patent 12580072
CLOUD ANALYTICS PACKAGES
5y 5m to grant Granted Mar 17, 2026
Patent 12555667
SYSTEMS AND METHODS FOR USING AI/ML AND FOR CARDIAC AND PULMONARY TREATMENT VIA AN ELECTROMECHANICAL MACHINE RELATED TO UROLOGIC DISORDERS AND ANTECEDENTS AND SEQUELAE OF CERTAIN UROLOGIC SURGERIES
2y 7m to grant Granted Feb 17, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

6-7
Expected OA Rounds
35%
Grant Probability
73%
With Interview (+37.7%)
High
PTA Risk
Based on 198 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month