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 January 15, 2026, is as follows:
Claims 1, 3, 5, 6, 7, 9, and 10, are pending.
Claim 2, 4 and 8 are canceled
The applicant has amended Claims 1, 3, 5 and 6 are amended and have been considered below
Request for Continued Examination
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on January 15, 2026 has been entered.
Response to Arguments
35 U.S.C. § 101 Rejection
Applicant’s arguments, see pages 5-8, filed January 15, 2026, with respect to amended Claims 1, 3, 5-7, and 9-10 have been fully considered and are not persuasive for the reasons stated below. The eligibility rejection under 35 U.S.C. § 101 is maintained.
Applicant Arguments Regarding Step 2A, Prong One (Abstract Idea)
The Applicant argues that the Examiner improperly digested the claims into an abstract "gist," and asserts that the claims do not recite a mental process because the "specific two-stage procedure where new patient data is assessed using clustering and then compared to historical data" using a prediction algorithm cannot be practically performed in the human mind or with pen and paper.
The Examiner respectfully disagrees and sustains the rejection. While the Examiner acknowledges the complexity of the "two-stage procedure," the claims are not rejected solely as "mental processes" but also as "Mathematical Concepts" under MPEP 2106.04(a)(2)(I). The complexity of a mathematical algorithm (e.g., K-means) does not remove it from the judicial exception of an abstract idea; rather, the recitation of the algorithm itself constitutes the exception regardless of whether a human can calculate it quickly. (Refer to further detail prong one analysis below.)
The Applicant argues that the claims merely "involve" an abstract idea (similar to a teeter-totter involving a lever) rather than "reciting" one because the claims incorporate specific clustering algorithms and specific medical features.
The Examiner respectfully disagrees and sustains the rejection because, the comparison to a physical teeter-totter is inapposite as the present claims lack any analogous physical structure and instead explicitly "recite" the mathematical formulas as the primary limitations of the method. Adding specific clinical variables (age, pain duration) to a mathematical model merely limits the abstract idea to a particular technological environment or "field of use," which does not save the claim from being directed to the exception under Prong One. (Refer to further detail prong one analysis below.)
The Applicant argues that Example 39 is a "perfect parallel" because it involves a two-stage approach for a neural network facial detector and concludes that such a system "does not recite any of the judicial exceptions."
The Examiner respectfully disagrees and sustains the rejection because, Example 39 is directed to a technical method of training a neural network to identify physical facial features to improve computer-vision functionality, whereas the present claims are directed to using a model to evaluate clinical data and output a numerical "likelihood" of medical success. Improving the accuracy of a mathematically calculated statistical prediction is considered an improvement to the abstract idea itself, not a technical improvement to a computer or another technology. (Refer to further detail prong one analysis below.)
Applicant Arguments Regarding Step 2A, Prong Two (Practical Application)
The Applicant argues that the amended claims as a whole reflect an "improvement in a technical field" and pass muster under Prong Two because the two-stage machine learning process overcomes a known technical shortcoming described in Goudman and Specification para. [0003].
The Examiner respectfully disagrees and sustains the rejection because the claims do not integrate the abstract idea into a practical application or qualify as a technological improvement. Applying "standard classification benchmarks" and "frequently used" algorithms like K-means (Spec. [0024], [0027]) via a generic server fails to fundamentally improve how the machine learning algorithm operates at a technical level (MPEP § 2106.05(a)). Furthermore, the claims fail to provide a practical medical application because they merely end at outputting a "predicted outcome representing a likelihood" without reciting a subsequent step that provided practicability for example actually administering a targeted neuromodulation treatment based on that prediction.
35 U.S.C. § 103 Rejection
Applicant's arguments, see pages 8–10, filed January 15, 2026, with respect to amended 1, 3, 5-7, and 9-10, have been fully considered and are not persuasive.
Note: The secondary reference has been updated from Goudman (cited in the prior Final Office Action, October 15, 2025) to Hoydonckx to more precisely address the amended claim language directed to neuromodulation treatment and baseline NRS/PCS clinical inputs. Applicant's arguments regarding the Goudman combination are therefore directed to a rejection that has been superseded and are moot with respect to the present grounds. The analysis below addresses the merits of Applicant's arguments as they apply to the current Neumann + Hoydonckx + Schnetz combination.
Applicant argues that Neumann uses the term 'cluster' strictly to describe a type of machine learning algorithm rather than to classify a patient into one of a plurality of predetermined clusters.
The Examiner respectfully disagrees and sustains the rejection. Under MPEP § 2145, an argument that attacks a single reference individually is not persuasive when the rejection rests on the teachings of a combination of references. The Examiner explicitly acknowledged Neumann's lack of a cluster-specific patient assignment mechanism and relied on Schnetz not Neumann to supply that limitation. Schnetz expressly teaches that 'a prognosis of one or more post-surgical outcomes of the test patient is determined based on the known surgical outcome of reference patients in the cluster including the test data vector' (Abstract; 0007, 0010, 0074,0122). This disclosure directly provides the claimed step of identifying a cluster from a plurality of clusters and confining the predictive output to reference patients within that specific cluster.
Applicant argues that in Neumann the cluster algorithm is the prediction algorithm, and that the prediction is selected based on data type rather than clustering results.
The Examiner respectfully disagrees and sustains the rejection for two independent reasons. First, under MPEP § 2145, Applicant improperly attacks Neumann individually for a limitation the Examiner explicitly mapped to Schnetz. The cluster-dependent prediction rule is supplied by Schnetz: 'determining a prognosis of the test patient based on the known post-surgical outcome of reference patients in the cluster including the test data vector' ( Abstract; 0007, 0010, 0074,00122). Second, Applicant mischaracterizes Neumann. Neumann explicitly teaches a multi-stage process in which 'clusters generated from the unsupervised machine-learning clustering algorithm may then be utilized in a supervised machine-learning algorithm' (Col. 7, ll. 15–25) a sequential two-stage pipeline directly contradicting Applicant's assertion that clustering and prediction are a single operation. Furthermore, Neumann discloses that the ameliorative learner evaluates data from a 'first training set 116' as well as a distinct 'second training set 152' (Col. 10, ll. 50–67; Col. 23, ll. 30–45), confirming the use of different data sets as the claim requires.
Applicant argues that the Examiner failed to identify where the prior art teaches a prediction stage dependent on the cluster stage outcome and the use of a different data set.
The Examiner respectfully disagrees and sustains the rejection. Under MPEP § 2145, Applicant attacks the primary reference in isolation, ignoring the express mapping of the cluster-dependent prediction to Schnetz. As established above, Schnetz 0122 explicitly provides the rule that the prognosis must be derived from reference patients confined to the identified cluster this is the cluster-dependent prediction the claim requires. With respect to the different data set, the Examiner explicitly identified Neumann's disclosure of an ameliorative learner that evaluates 'data from first training set 116 as well as data from second training set 152' (Col. 10, ll. 50–67), directly mapping the use of multiple distinct data sets. These are specific textual citations not conclusory statements and they satisfy the articulated reasoning requirement of KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398, 418 (2007) and MPEP § 2141(III).
Applicant argues that the Examiner's rejection consists of conclusory statements without articulated reasoning sufficient to satisfy KSR, alleging the Examiner merely generalized the prior art.
The Examiner respectfully disagrees and sustains the rejection. The rejection is not conclusory. The Examiner provided explicit column-and-line citations to Neumann (Col. 7, ll. 15–25 for two-stage pipeline; Col. 10, ll. 50–67 for first training set; Col. 23, ll. 30–45 for second training set), paragraph citations to Schnetz (abstract, 0010, 0007, 0051, 0066, 0122, 0176 for cluster-specific prognosis), and page-specific citations to Hoydonckx (p. 34 for SCS neuromodulation; p. 46 for NRS/PCS baseline domains). These are not generalizations they are documentary evidence of specific disclosures. In addition, the Examiner articulated three KSR-compliant motivations to combine: (1) both Neumann and Hoydonckx address patients refractory to conventional management, creating a shared objective; (2) Neumann identifies inaccurate treatment selection as its primary problem (Col. 1, ll. 25–30), and Schnetz's cluster-conditioned prognosis resolves that exact gap; and (3) Schnetz's method is architecturally compatible with Neumann's existing pipeline, confirming a reasonable expectation of success without undue experimentation. Under KSR and MPEP § 2141(III), this is sufficient. Applicant's repeated focus on Neumann in isolation fails to traverse the § 103 combination under MPEP § 2145.
Claim Rejections – 35 USC 112(d)
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claim 9-10 rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends.
Claim 9 is indefinite because it explicitly depends on Claim 8 ("The method of claim 8"), but Claim 8 has been canceled, leaving Claim 9 without a valid base claim to provide its foundational limitations. Claim 10 is also rejected because dependent to claim 9, but claim 9 depends on claims 8 that is canceled. For purposes of examination, claim 9 will be interpreted as depending on claim 6.
Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
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.
Subject Matter Eligibility Analysis:
Claims 1, 3, 5-7, and 9-10 are rejected under 35 U.S.C. § 101 because the claimed subject matter is directed to a judicial exception (an abstract idea) without reciting elements that integrate the exception into a practical application or provide an inventive concept amounting to significantly more than the exception itself.
Step 1: Statutory Categories Analysis
The claims fall within the statutory categories of invention.
Process (Claims 6-7, 9-10): The language reciting "collecting a first plurality of patient features... using a machine learning engine to apply a K-means algorithm... to output a predicted outcome" defines a series of acts or steps, fulfilling the definition of a process in MPEP § 2106.03.
Machine (Claims 1, 3, 5): The language reciting "A system for predicting an outcome... comprising: a server... and a machine learning engine" describes a concrete thing consisting of parts, fulfilling the definition of a machine in MPEP § 2106.03.
Step 2A, Prong One: Judicial Exception Analysis
Step 2A, Prong One determines whether the claims are directed to a judicial exception, such as an abstract idea, under MPEP 2106.04.
The whole invention is related to an approach that uses machine learning predictive modeling to predict patient response to spinal cord stimulation treatment based on historical patient characteristics. Refer to Spec., para. [0004], [0005], and Figure 3 for further details.
More specify, claims 1, 3, 5-7, and 9-10 are directed to an abstract idea, specifically mathematical concepts and mental processes. The invention focuses on an approach that uses machine learning predictive modeling to forecast patient response to spinal cord stimulation treatment based on historical patient characteristics (Spec., para. [0004], [0005], Figure 3). Claims 1 and 6 recite receiving clinical patient data sets, mathematically grouping the patient into a cluster using a K-means algorithm, and applying a mathematical predictive model corresponding to that cluster to output a likelihood of treatment success using a second set of data.
Independent Claim 1 Recites the following non-bold parts abstract idea:
A system for predicting an outcome of a neuromodulation treatment, comprising:
a server providing a user interface configured to accept a first set of data representing a first plurality of features from a new patient for whom a prediction of spinal cord stimulation is desired, wherein the first plurality of features include at least a patient age, a pain duration, a baseline NRS score, and a baseline PCS score, and to accept a second set of data representing a second plurality of features that is different than the first set of features;
and a machine learning engine in communication with the server, wherein the machine learning engine is configured to apply a K-means algorithm to perform a cluster stage to evaluate the plurality of patient features to identify a cluster from a plurality of clusters that corresponds to the plurality of features of the new patient based on a first set of data from a set of patients with known outcomes and then to perform a prediction stage using a predictive model selected from the group consisting of logistic regression, random forest, XGBoost, elasticnet, support vector machine, Naïve Bayes, and combinations thereof that provides a predicted outcome representing a likelihood that the neuromodulation treatment will produce a positive response for the new patient by selecting the predictive model that is specific to the cluster identified in the cluster stage and then applying the predictive model that is specific to the cluster identified by the cluster stage to a second set of data from the set of patients with known outcomes that is different than the first set of data.
Claim Abstract Classification Rationale
Under their Broadest Reasonable Interpretation (MPEP § 2111), the independent claims recite gathering patient data, executing mathematical formulas to group the data, and performing mathematical calculations to output a probability.
Mathematical Concepts (MPEP § 2106.04(a)(2)(I)):
The claims recite mathematical concepts because they rely on mathematical relationships, formulas, and calculations. Independent claims 1 and 6 recite applying "a K-means algorithm to perform a cluster stage", a mathematical formula that minimizes within-cluster sum of squares to group data points, and applying "a predictive model selected from the group consisting of logistic regression, random forest, XGBoost, elasticnet, support vector machine, Naïve Bayes" each of which is a mathematical calculation that computes a probability output. These mathematical formulas receive numerical patient data as inputs and produce a numerical likelihood as output.
Mental Process (MPEP § 2106.04(a)(2)(III)):
At a higher level of abstraction, the claims also recite a mental process. Independent claims 1 and 6 recite "evaluate the plurality of patient features to identify a cluster" and "provides a predicted outcome representing a likelihood." The specification frames the invention as a tool to "provide an objective datapoint to augment the clinician's decision about when to pursue alternate therapies" (Spec., para. [0002]). The evaluation of patient characteristics to identify a patient subgroup and the judgment of likely treatment outcomes are cognitive steps, observation, evaluation, and opinion that a clinician performs when assessing whether a patient is a good candidate for spinal cord stimulation.
Manual Replication Scenario (Human Equivalence)
A clinician gathers the patient's age, pain duration, NRS, and PCS scores. The clinician reviews historical patient profiles and, based on experience, mentally groups the patient with similar past patients. The clinician then recalls the success rates observed for that patient group and forms a judgment about the likelihood of treatment success. The claims merely automate these fundamental observation-grouping-prediction steps by substituting the clinician's mental heuristics with specific mathematical formulas (like K-means and logistic regression).
Dependent Claims Analysis
The dependent claims are also directed to the abstract ideas of mathematical concepts and mental processes.
Claims 3, 7, and 9: Claim 3 recites a "machine learning algorithm trained with data representing the plurality of features." Claim 7 recites "K-means clustering of data." Claim 9 recites algorithms "selected from the group consisting of logistic regression, random forest, XGBoost, elasticnet, support vector machine, Naïve Bayes." These claims identify specific mathematical formulas and statistical training iterations, falling strictly under Mathematical Concepts.
Claims 5 and 10: Claim 5 recites the second plurality of features are "selected from at least one of demographics, pain descriptors, pain questionnaire data, psychiatric comorbidities, spinal imaging, activity, medications, non-psychiatric comorbidities, and past spinal cord stimulation results." Claim 10 recites identical language. These claims identify specific categories of medical data, falling under Mental Process / Data Collection.
Because the claims are directed to an abstract idea, the analysis proceeds to Step 2A, Prong Two to determine if it is integrated into a practical application.
Step 2A, Prong Two: Integration into a Practical Application
The claims do not integrate the abstract idea into a practical application. The additional elements simply provide a generic technological environment to perform the data analysis and fail to impose meaningful limits on the abstract idea.
Evaluation of Independent Claims 1 and 6 Additional Elements Generic Hardware (server and user interface): The recitation of a server and user interface is a mere instruction to implement the abstract idea on a generic computer (MPEP § 2106.05(f)). The server and interface are invoked solely to "accept a first set of data" and "accept a second set of data." Obtaining data inputs for mathematical calculations is insignificant extra-solution activity (MPEP § 2106.05(g)). The claims do not recite any structural modifications that improve the functioning of the server or user interface itself (MPEP § 2106.05(a)).
Generic Software and Data Architecture (machine learning engine):
The claims require the machine learning engine to accept a "first set of data" for the cluster stage and a "second set of data" for the prediction stage. This two-stage data architecture reflects a generic mathematical data-organization methodology, using different feature subsets for clustering versus classification to prevent overfitting, rather than a technological improvement to the computer's functioning (MPEP § 2106.05(a)). The separation of data inputs is an inherent part of the mathematical modeling, not a structural enhancement of the machine learning engine or the server. Reciting the engine in the context of "predicting an outcome of a neuromodulation treatment" merely links the mathematical abstract idea to a particular technological environment (MPEP § 2106.05(h)).
Combination as a Whole:
The specific ordered combination of a server accepting data, grouping it via K-means clustering, performing a cluster-specific prediction using a second data set, and outputting the result does not transform the abstract idea. The sequence is the logical execution of the mathematical methodology itself. The elements operate in their generic capacities—the server receives data, and the processor calculates the formulas, without demonstrating any technical synergy that improves the underlying computing system.
Dependent Claims Analysis The dependent claims do not add new technical additional elements; they merely narrow the abstract idea.
Claims 3, 7, and 9: Reciting specific algorithms ("logistic regression, random forest," etc.) or training requirements fails to improve computer functionality (MPEP § 2106.05(a)) because it simply identifies the specific mathematical formula to be calculated by the generic processor.
Claims 5 and 10: Reciting specific medical data variables ("demographics, pain descriptors," etc.) is a mere field-of-use limitation (MPEP § 2106.05(h)) that restricts the data gathering to the medical field without providing a practical, technical application.
When viewed as a whole, the combination of these elements in the dependent and independent claims does not integrate the abstract idea into a practical application because the claims merely direct the application of statistical algorithms to specific medical data using a generically invoked server.
Step 2B: Inventive Concept Analysis
The claims lack an inventive concept because the additional elements, alone and in combination, represent well-understood, routine, and conventional activities in the field that do not amount to significantly more than the abstract idea itself.
Evaluation of Independent Claims 1 and 6 Additional Elements Generic Hardware (server and user interface):
The invocation of a server and user interface provides no inventive concept. The specification admits these components are well-understood, routine, and conventional off-the-shelf components. The specification explicitly states that "GUI files will be located and loaded from a server 16, such as Amazon web services (AWS)" and "As is known in the art, GUIs can require user authentication and login..." (Spec., para. [0022]). This is a mere instruction to apply the exception using generic commercial cloud components (MPEP § 2106.05(f)).
Generic Software (machine learning engine):
Applying existing machine learning algorithms to evaluate data is well-understood, routine, and conventional in the field of data science. The specification admits "The K-means algorithm is one of the simplest and most frequently used clustering algorithms" (Spec., para. [0024]) and identifies the predictive models as generic tools like "logistic regression, random forest, XGBoost" (Spec., para. [0005]). Utilizing "frequently used" algorithms on a generic engine does not constitute a technological improvement (MPEP § 2106.05(a)). The two-stage data feature separation is likewise a conventional data science technique to optimize mathematical model performance. Refer also to Mars, US10296848, Col. 6, ll. 1 – 50, Charles, US20190108912A, par. 0017, 0063, 0053, Mamta, US20190325354A1, par. 0031, 0052-0053
Combination as a Whole:
The ordered combination of using a generic web server to receive patient data and passing it to a conventional machine learning engine executing generic K-means and logistic regression formulas is a well-understood, routine, and conventional arrangement. The whole is no greater than the sum of its generic parts, merely automating an abstract idea.
Dependent Claims Analysis The dependent claims do not introduce an inventive concept.
Claims 3, 7, and 9: These claims add specific mathematical algorithms ("XGBoost, elasticnet," etc.). This is MPEP § 2106.05(f) - Mere Instructions to apply the exception using specific formulas. The specification confirms this is a generic selection from known options: "The machine learning algorithm may comprise logistic regression, random forest, XGBoost... or combinations thereof." (Spec., para. [0005]).
Claims 5 and 10: These claims add types of medical data ("psychiatric comorbidities, spinal imaging," etc.), which is insignificant pre-solution activity (g) and a mere field-of-use limitation (h). The specification confirms these are generic clinical data inputs for assessing neuromodulation (Spec., para. [0021]).
As a whole, the combination of the dependent and independent claims merely automates the abstract idea of medical outcome prediction using generic, off-the-shelf cloud computing and generic statistical methodologies.
Therefore, Claims 1, 3, 5-7, and 9-10 are rejected under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, 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.
Claim(s) 1, 3, 5-7, and 9-10 are rejected under 35 U.S.C. 103 as being unpatentable over US11929170- Neumann, and Hoydonckx, Y., Costanzi, M., & Bhatia, A. (2019). A scoping review of novel spinal cord stimulation modes for complex regional pain syndrome. Canadian Journal of Pain, 3(1), 33-48. doi.org.- refer to PTO-892 letter U and in view of US20190046122A1-Schnetz.
Neumann teaches Claim 1.
A system for predicting an outcome of
a server providing a user interface configured to accept data representing a plurality of features from a new patient for whom a prediction of ; (Neumann, abstract, (Column 20, lines 49- 65), (Column 26, lines 46-56), (Column 35 lines24-30, (Column 27, line 4-15), (Column 1, lines 20-42), (Column 1, line 45-55) Neumann discloses receiving user input via mechanisms like "a form or similar graphical user interface object" or "questions and/or surveys based on user input generated at user client device 140", fulfilling the BRI of a "user interface." Neumann further details acceptance of varied patient-specific data ("plurality of features") including results from" may include a medical report table 316, which may list textual descriptions of medical tests..." and "user life element datums". The purpose of receiving this data is for generating "prognostic outputs" and selecting "ameliorative outputs", which aligns with the BRI of using the data for a "prediction" related to a medical treatment.
Neumann further describes a server receiving a user life element datum, prognostic output.
wherein the first plurality of features include at least a patient age, a pain duration, ; (Neumann, See at least, Col. 44, ll. 5-15, group of people having a shared age range ... Col. 12, ll. 49-67, evaluations of sensor, including measures of audition, vision, olfaction, gustation, vestibular function and pain ... first training set 116 including a plurality of first data entries ... Col. 23, ll. 30-45, second training set 152 including a plurality of second data entries.. Col. 14, ll. 14-25, limitation current or past physical…) Neumann is read on patient age, pain duration, and receiving different sets of data because the prior art describes collecting demographic age ranges, evaluating pain, and processing first and second distinct data entries.
and a machine learning engine in communication with the server, wherein the machine learning engine is configured to(Neumann, abstract, Column 9, line 55 - 67, Column 10, line 1 -25) Neumann describes a communication between a server and machine learning, as it states that "at least a server 104 and/or prognostic label learner 108 may be configured to select training data to generate prognostic output using a selected machine-learning process." This indicates the server's role in providing data for the machine learning process to produce results.
a cluster stage to evaluate the plurality of patient features to identify a cluster from a plurality of clusters that corresponds the plurality of features of the new patient based on a first set of data from a set of patients with known outcomes and then to perform a prediction stage predictive model selected from the group consisting of logistic regression, random forest, XGBoost, elasticnet, support vector machine, Naïve Bayes, and combinations thereof that provides a predicted outcome representing a likelihood that the positive response for the new patient by selecting the predictive model tha (Neumann, (Column 5, lines 59-67, Column 6, lines 1-50), (Column 7, lines 1-25), (Column 8, lines 24-50), (Column 9, lines 1-30, 50-67), Column 26, lines 18 - 35), (Column 25, line 43-67), (Column 9, line 1- line 25), figure 1, figure 2, (Column 22, line 1- line 35), (Column 6, line 1 - line 15), (Column 43, lines 1-40), Column 10, lines 5–67, Column 24, lines 15-30, Column 44, lines 10-25, Col. 10, ll. 50-67 )
Neumann anticipates the cluster stage, specific predictive models, and application to a second data set by describing a multi-stage machine learning system that groups historical patient data to dictate the selection of a subsequent supervised learning formula. Specifically, the prior art discloses utilizing an "unsupervised machine-learning clustering algorithm" to generate data clusters that "may then be utilized in a supervised machine-learning algorithm." Furthermore, Neumann explicitly anticipates the claimed prediction stage algorithms by naming an "elastic net model," "logistic regression model," "support vector machines," "naïve Bayes methods," and a "forest of randomized tress" to determine the "probability of treatment success." Finally, the prior art maps to applying the cluster-specific model to different data by disclosing an ameliorative learner that evaluates "data from first training set 116 as well as data from second training set 152" to output these treatment predictions.
Neumann describes a “prediction stage” that uses a trained model to translate prognostic likelihood scores into treatment-focused predictions, while the ameliorative learner creates a model using a second training set to assess treatment success probabilities and prognostic improvement scores.Obvious Rationale:
Neumann disclosed all elements except the element's strikethroughs, that are explain by Hoydonckx and Schnetz.
neuromodulation treatment and spinal cord stimulation:
Hoydonckx teaches neuromodulation treatment and spinal cord stimulation of the Claim 1. (Reference Hoydonckx, See at least page 34, Neuromodulation approaches including spinal cord, dorsal root ganglia, and peripheral nerve stimulation have been used for chronic pain syndromes and evaluate the analgesic impact and adverse effects of PF-SCS on patients).
The combination of Neumann + Hoydonckx applications make obvious the full limitation under a predicted outcome representing a likelihood that the neuromodulation treatment will produce a positive response for the new patient because a POSITA would implement Hoydonckx's spinal cord stimulation as the specific treatment target within Neumann's predictive workflow. Since Neumann's server is configured to generate success scores by utilizing a "prognostic label learner" to create a "prognostic machine-learning model" that calculates a "prognostic probability score" based on "prevalence and predictive values," and Hoydonckx identifies spinal cord stimulation as a high-efficacy "neuromodulatory intervention," a POSITA would predictably achieve the claimed forecasting by populating Neumann's learning model with Hoydonckx's clinical outcomes. This integration yields the predictable result of a success-likelihood score for a stimulation treatment. (Reference, See at least, Neumann states prognostic label learner 108 is designed and configured to generate at least a prognostic output using machine-learning processes (Col. 5, ll. 67 - Col. 6, ll. 8); and predictive value indicates the probability of a prognosis in an individual... [and] may be calculated based on prevalence and predictive values and prognostic label learner may generate a plurality of prognostic outputs which may each be ranked... based on a decreasing likelihood (Col. 9, ll. 1-44)).
A skilled Artisan in the art who read Neumann application, would combine Hoydonckx with Neumann, because both are in the same subject matter of medical treatment where Neumann focuses on selecting an ameliorative output using artificial intelligence and Hoydonckx provides a review of novel spinal cord stimulation modes. (Reference, See at least, Neumann states methods and systems for selecting an ameliorative output using artificial intelligence and seeks to help patients who are unresponsive to conventional medical management; and Hoydonckx states spinal cord stimulation (PB-SCS)... page 35->Population: Studies included in the clinical analy-sis focused on adult patients (at least 18 years of age), refractive to conventional medical manage-ment and or conventional PB-SCS..)
A skilled artisan would be motivated to combine Neumann and Hoydonckx because integrating the missing neuromodulation treatment and spinal cord stimulation therapies into Neumann's artificial intelligence framework resolves the primary art's identified problem of inaccurate treatment selection. Since Neumann explicitly seeks to prevent unsuccessful medical consequences by generating targeted ameliorative outputs, and Hoydonckx explicitly evaluates the analgesic impact of spinal cord stimulation to treat refractory pain, merging these teachings provides a distinct clinical benefit. The benefit of this integration is the ability to preemptively accurately forecast a positive response to spinal cord stimulation before a patient undergoes the invasive procedure, yielding the expected predictive result of a data-driven predicted outcome for neuromodulation success. Reference, See at least, Neumann states Inaccurate selection can ultimately frustrate users and lead to unsuccessful consequences (Col. 1, ll. 10-30); and Hoydonckx states failure to relieve pain and resolve associated symptoms of CRPS is common with these pharmacological and physical therapies. Neuromodulation approaches including spinal cord... stimulation have been used (p. 34) and evaluate the analgesic impact and adverse effects of PF-SCS on patients (p. 35).
A skilled artisan would have a reasonable expectation of success in combining these references as a whole because Neumann provides a computationally agnostic machine-learning architecture that is already fully configured to accept diverse physiological datasets to output probability scores.
baseline NRS score and a baseline PCS score:Hoydonckx teaches a baseline NRS score and a baseline PCS score of Claim 1, ( Hoydonckx, See at least page 46, Domains to be assessed at baseline...Pain-related: NRS...Psychological status, e.g., PCS; Pre-intervention: NRS pain: 8.2... page 38, PCS 33; and page 45, the importance of evaluating pain-related domains in clinical and research settings is established and validated tools to evaluate these domains are also widely available.
The combination of Neumann + Hoydonckx makes obvious, a baseline NRS score, and a baseline PCS score because a POSITA reading Neumann's requirement for numerical score data entered by an evaluating professional and/or by a subject performing a self-test such as a computerized questionnaire would implement Hoydonckx's NRS and PCS as the validated baseline instruments to supply those inputs, Neumann seeks standardized, quantifiable pain-domain data to train its prognostic engine, and Hoydonckx teaches the specific validated technique that predictably achieves that goal on the same patient input type, with the predictable result of a machine learning model trained on uniform, clinically reliable baseline data. (Neumann, See at least, Col.12, ll. 30-43 numerical score data entered by an evaluating professional and/or by a subject performing a self-test such as a computerized questionnaire; Hoydonckx, See at least, page 45, Domains to be assessed at baseline...Pain-related: NRS...Psychological status, e.g., PCS; Pre-intervention: NRS pain: 8.2...PCS 33 …validated tools to evaluate these domains are also widely available…)
A skilled artisan who read Neumann's application would combine Hoydonckx with Neumann because Neumann presents a specific problem, its machine learning engine requires standardized, validated clinical scale inputs for pain-domain patient populations but does not name which scales to use and Hoydonckx directly resolves that gap by identifying NRS and PCS as the clinically established instruments for exactly that domain, making the combination a natural and directed step rather than a speculative one.
A POSITA would be motivated to combine Neumann + Hoydonckx because integrating Hoydonckx's NRS and PCS into Neumann's intake workflow resolves Neumann's unspecified input problem replacing variable user symptom descriptions with validated, reproducible baseline metrics yielding the specific predictable benefit of a more clinically consistent and reproducible prognostic output from Neumann's machine learning engine without modifying its architecture.
A POSITA would have a reasonable expectation of success because Neumann's architecture already accepts numerical score inputs from clinical self-assessments and professional evaluations NRS and PCS are architecturally compatible instruments requiring no structural redesign of Neumann's system and Hoydonckx documents their established reliability and availability in exactly the clinical context Neumann targets, confirming that substituting these validated tools into Neumann's data intake pathway produces standardized baseline pain data without undue experimentation.
K-means algorithm
Neumann, + Hoydonckx for SCS prediction, teaches a machine learning system suitable for predicting medical treatment outcomes, including "selecting an ameliorative output using artificial intelligence" (Neumann, Abstract). This system utilizes a machine learning engine (Neumann, Column 20, lines 49-65) employing diverse patient features (Neumann, Column 19, 55-67, Column 20, lines 1-10,) and incorporating processes which "may cluster data" (Neumann, Column 43, lines 10-20). Thus, Neumann provides a cluster stage.
However, Neumann, + Hoydonckx lacks to explicitly identify K-means clustering as the required algorithm. While Neumann teaches using data with "known outcomes" (like "first prognostic label 124" (Neumann, Column 10, lines 52-67) and "second prognostic label" (Neumann, Column 23, lines 31-64)) to train predictive models, it does not explicitly teach defining the clusters themselves based on data from patients with known outcomes using K-means, as the claim requires.
Schnetz provides the specific, missing elements. Schnetz teaches determining patient prognosis via clustering and explicitly discloses the step of performing a “K-means clustering procedure” (Schnetz, 0007, 0010, abstract, Claim 1). Furthermore, Schnetz explicitly teaches defining these clusters by using “reference data vectors... from reference patients with known clinical outcome” (Schnetz, Abstract, abstract, 0010, 0007, 0051, 0066, 0122, 0176), doing so for the specific purpose of “determining a prognosis” (Schnetz, Abstract, 0007, 0010, 0122).
A person of ordinary skill in the art (POSITA), seeking to improve the Neumann/ Hoydonckx predictive system by refining its unspecified clustering capability, would have been motivated to incorporate the specific clustering method taught by Schnetz. Schnetz offers a known technique (K-means using known outcome data) demonstrated for the highly relevant purpose of “determining a prognosis” (Schnetz, Abstract, 0007, 0010) through patient data clustering. Applying Schnetz's specific method is a predictable way to enhance the patient stratification in Neumann's system, leveraging Schnetz’s teaching that K-means clustering allows for automated patient data grouping based on similarities (Schnetz, 0051, 0066) to improve the clinical relevance for SCS outcome prediction.
A POSITA would have had a reasonable expectation of success. K-means is a standard algorithm suitable for the type of patient data used by Neumann (Neumann, 33-39, 52, 68). Schnetz demonstrates this specific approach K-means defined by known outcomes yields useful clusters for medical prognosis (Schnetz, Abstract, 0007, 0010, 0066, 0122), confirming its feasibility in a similar context.
Selecting the Predictive Model Specific to the Cluster Identified in the Cluster Stage
Schnetz teaches a cluster-specific prognosis determination of Claim 1, selecting the predictive model that is specific to the cluster identified in the cluster stage, that required applying a prediction that is bound to the particular cluster to which the test patient was assigned, by explicitly conditioning the prognosis output on the identity of the cluster containing the test data vector meaning the predictive result is not global but is uniquely derived from the reference patients confined to that cluster.
(Schnetz, See at least, A prognosis of one or more post-surgical outcomes of the test patient is determined based on the known surgical outcome of reference patients in the cluster including the test data vector (Abstract); determining a prognosis of the test patient based on the known post-surgical outcome of reference patients in the cluster including the test data vector (abstract, 0010, 0007, 0051, 0066, 0122, 0176); the reference data vectors characterize concurrent MAP, MAC, and BIS measures for sequential time intervals during surgical procedures of reference patients with a known post-surgical outcome; and the method further comprises determining a prognosis of the test patient based on the known post-surgical outcome of reference patients in the cluster including the test data vector (0074)).
The combination of Neumann + Schnetz makes obvious the full limitation under selecting the predictive model that is specific to the cluster identified in the cluster stage because both Neumann and Schnetz operate in the same domain of machine-learning-driven medical prognosis using clustered patient data. Neumann supplies the multi-stage ML pipeline including unsupervised clustering feeding a supervised model while Schnetz supplies the specific architectural rule that the predictive output must be governed by the cluster to which the patient belongs. POSITA would Combining these teachings produces a predictable outcome: a pipeline in which the supervised model applied in the prediction stage is selected based on the cluster identity, precisely the arrangement the claim recites. (See at least, Neumann: clusters generated from the unsupervised machine-learning algorithm may then be utilized in a supervised machine-learning algorithm (Col. 7, ll. 1–25); Schnetz: A prognosis of one or more post-surgical outcomes of the test patient is determined based on the known surgical outcome of reference patients in the cluster including the test data vector (Abstract)).
A skilled artisan who read Neumann's application would combine Schnetz with Neumann because Neumann identifies a specific problem with treatment selection that is directly resolved by Schnetz. Neumann addresses the challenge that inaccurate medical forecasting can lead to failed treatments. To improve accuracy, Neumann teaches a multi-stage system where an unsupervised clustering algorithm groups patient data, which is then fed into a supervised model to generate a prediction. However, Neumann does not provide the specific rule for how the supervised model should leverage those clusters for the final prognosis. Schnetz resolves this exact gap by teaching that a reliable prognosis must be determined directly from the known outcomes of the specific reference patients located inside the identified cluster. (Reference, See at least, Neumann: "Accurate selection of ameliorative outputs can be challenging. ... Inaccurate selection can ultimately frustrate users and lead to unsuccessful consequences." (Col. 1, ll. 25-30) and "clusters generated from the unsupervised machine-learning algorithm may then be utilized in a supervised machine-learning algorithm" (Col. 7, ll. 15-25); Schnetz: "A prognosis of one or more post-surgical outcomes of the test patient is determined based on the known surgical outcome of reference patients in the cluster including the test data vector" (Abstract; 0122)).
A skilled artisan would be motivated to combine Neumann and Schnetz because integrating Schnetz’s cluster-specific rule into Neumann’s framework provides the exact missing criteria needed to solve Neumann's problem of inaccurate treatment forecasting. By adopting Schnetz's teaching to explicitly bind the predictive outcome to the specific cluster a patient belongs to, the system eliminates generalized, less accurate predictions. The benefit of this integration is the ability to generate a highly tailored and accurate prognostic output for each patient subgroup, yielding the predictable result of preventing unsuccessful medical treatments through precise, data-driven forecasting.
(Reference, See at least, Neumann: "Accurate selection of ameliorative outputs can be challenging... Inaccurate selection can ultimately frustrate users and lead to unsuccessful consequences." (Col. 1, ll. 25-30); Schnetz: "determining a prognosis of the test patient based on the known post-surgical outcome of reference patients in the cluster including the test data vector" (0122)).
A skilled artisan would have a reasonable expectation of success in combining these teachings as a whole because Neumann’s architecture is already explicitly built to accept outputs from a clustering algorithm and feed them into a downstream supervised predictive model. Schnetz’s cluster-conditioned prediction rule is structurally and functionally compatible with this exact pipeline. Because Schnetz demonstrates that confining a prognosis to a specific reference cluster yields clinically meaningful results in a validated medical context, an artisan would predictably understand that applying this identical rule within Neumann’s compatible framework would operate successfully without requiring any structural redesign or undue experimentation.
Note: Claim 6 recites substantially similar subject matter as claim 1, and is also rejected under the same analysis.
Neumman and Hoydonckx in further view of Schnetz teaches Claim 3.
The system of claim 1, wherein the predictive model comprises a machine learning algorithm trained with data representing the plurality of features from the set of patients having known outcomes.
Neumann explicitly discloses creating and using a "prognostic machine-learning model 112" Neumann, (Column 6, lines 14-36) and an "ameliorative machine-learning model 148" Neumann, Column 22, lines 24-40, which comprise machine learning algorithms Neumann. Neumann further explicitly teaches that these models are trained using a "first training set" Neumann, Paragraph 26 containing entries with patient features ("physiological state data 120") correlated with known outcomes ("first prognostic label 124") Neumann, Column 10 lines 52-67, and a "second training set 152" Neumann, column 23, lines 32-66, containing known outcomes ("second prognostic label") correlated with treatment labels.
Neumman and Hoydonckx in further view of Schnetz teaches Claim 5.
The system of claim 1, wherein the second plurality of features are selected from at least one of demographics, pain descriptors, pain questionnaire data, psychiatric comorbidities, spinal imaging, activity, medications, non-psychiatric comorbidities, and past spinal cord stimulation results.
Neumann discloses accepting and using a wide variety of patient features that fall into multiple categories listed in the claim, including "measures of physical capability such as... grip strength... gait speed" Column 11, lines 57-67, column 12, lines 1-10 (Activity), "measures of psychological function or state" Column 12, lines 9-26 (Psychiatric comorbidities), "user responses to questions on a psychological, behavioral, personality, or cognitive test" Column 20, lines 54-67 (Questionnaire data, Psychiatric comorbidities), user descriptions of "symptoms" and past "pharmaceuticals" Paragraph 39 (Pain descriptors, Medications), and various physiological data Column 13, lines 14-36 (Non-psychiatric comorbidities).
Claims 7, 9, and 10 recite substantially similar subject matter as claims 3 and 5, and are also rejected under the same analysis.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOSHUA DAMIAN RUIZ whose telephone number is (571)272-0409. The examiner can normally be reached 0800-1800.
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 at (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.
/JOSHUA DAMIAN RUIZ/Examiner, Art Unit 3684
/Shahid Merchant/Supervisory Patent Examiner, Art Unit 3684