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 .
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 7/10/25 are in accordance with the provisions of 37 CFR 1.97 and are considered by the Examiner.
Priority
Claim FOR app# CN202310752854.4 priority and PTC/ CN2024/093315 are acknowledged.
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier.
Claim 17, terms, “vital sign data acquisition module,” “multimodal data acquisition module,” and “physical state data generating module” invoke 35 U.S.C. 112(f) interpretation.
Claim 17.
…
a vital sign data acquisition module, configured to …
a multimodal data acquisition module, configured to …
a physical state data generating module, configured to ….
Claim 18, terms, “physical state data acquisition module” and “health assessment module” invoke 35 U.S.C. 112(f) interpretation.
Claim 18.
A health assessment device, comprising:
a physical state data acquisition module, configured to … the data processing method according to claims 1;
a health assessment module, configured to …
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
(Claim 17) “vital sign data acquisition module”
Under §112(f), this module is interpreted be a component of a processing device as shown in Fig. 7 & [0141]-[0149] & [0183]-[0184] that receives sensor data from separate sensor components and analyzes the data.
(Claim 17) “multimodal data acquisition module”
Under §112(f), this module is interpreted as be a component of a processing device as shown in Fig. 7 & [0143] & [0183]-[0184] that receives imaging or survey data from separate imaging and/or survey input components and analyzes the data.
(Claim 17) “physical state data generating module”
Under §112(f), this module is interpreted be a component of a processing device as shown in Fig. 7 & [0144] & [0183]-[0184] that performs some kind of analysis. please make it clearer how you are interpreting this module
(Claim 18 )“health assessment module”
The specification provides explicit software submodules/algorithms and a hardware execution context: it states the “health assessment module comprises: a resampling submodule… a difference processing submodule… the input submodule” ([0171]–[0173]). It also discloses concrete model structures used by that module, e.g., “ARIMA… Informer… N-BeatXs” and “the fifth model is a LightGBM model” ([0031]–[0043]). Finally, it anchors execution to hardware by describing an electronic device with “a memory, a processor, and a program… executable on the processor” ([0051]), so the corresponding structure is software algorithms running on processor-based hardware.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
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 3,4, 8-16 and 18 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 3
Term: “the target disease” (antecedent basis)
Issue: Claim 3 recites “degree of impact … on the target disease,” but claims 1-2 do not previously introduce any “target disease,” making the referent unclear as drafted. Under MPEP, a claim is indefinite where it is unclear what earlier element “the/said” language refers to (e.g., “said lever” with no prior lever).
Claim 8 and 18
Term: “physical condition data” (lack of antecedent basis / inconsistency)
Issue: Claim 8 first recites “acquiring physical state data,” but then requires “inputting the physical condition data into a health assessment model,” without introducing “physical condition data” or clarifying whether it is the same as “physical state data.” Claim 18 is also found to be indefinite under the same analysis
Claim 14
Term: “third model and the third model” (unclear reference / internal inconsistency)
Issue: Claim 14 recites “splicing output results of the first model, the second model, the third model and the third model,” which is internally inconsistent and leaves it unclear which distinct model outputs are included/excluded (e.g., whether a “fourth model” was intended).
Claim 9
Term: “the resampled vital sign data”
Issue: Antecedent basis rejection; none of the parent claims describe resampling vital sign data.
Note: Claims 4, which depends on claim 3, and claims 9–13 and 14–16, which depend on claim 8, are rejected because dependent claims inherit the language of the claims from which they depend.
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.
Non-Statutory Rejection Rational:
Claim 20 recites a “readable storage medium, storing a program,” which is not directed to a tangible article (a manufacture) because it is a signal-per se, storage medium that could be interpreted as a transitory signals which are not patent eligible, and the specification does not limit the definition to non-transitory medium. In the interest of compact prosecution for further eligibility analysis purposes, Examiner interprets claim 20 as being directed to the statutory subject matter of a non-transitory computer-readable medium. Refer spec. par. 0180 – 0182.
Subject Matter eligibility Rejection 35 U.S.C 101
Claims 1-20 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 (35 U.S.C. §101)
The claims do fall within at least one of the four statutory categories (process, machine, manufacture).
Process (Claims 1–16)
Claims 1–16 recite methods (a series of acts/steps), e.g., Claim 1 recites “acquiring vital sign data… collecting multimodal data… generating physical state data,” and the dependent method claims further recite steps such as “binning,” “grouping,” “generating supplementary data,” “standardizing,” and “inputting… into a health assessment model.” These are classic “process” limitations under Step 1.
Machine (Claims 17–19)
Claims 17–19 recite devices with components, e.g., Claim 17 “A data processing device, comprising: a vital sign data acquisition module… a multimodal data acquisition module… a physical state data generating module,” Claim 18 “A health assessment device, comprising…,” and Claim 19 “An electronic device, comprising: a memory, a processor, and a program….” This language describes “a concrete thing consisting of parts,” i.e., a machine for Step 1.
Manufacture (Claim 20)
Claim 20 recites a “readable storage medium, storing a program,” which is interpreted as directed to a tangible article (a manufacture) in the purpose of compact prosecution and future applicant amendment.
The analysis proceeds to Step 2A Prong One for compact prosecution.
2A, Prong One:
Step 2A Prong One asks whether the claim recites a judicial exception (law of nature, natural phenomenon, or abstract idea). Under USPTO guidance, abstract ideas are analyzed using the groupings in MPEP §2106.04(a)(2) (mathematical concepts, certain methods of organizing human activity, and mental processes).
The invention is directed to collecting vital-sign and multimodal patient data (images/surveys), generating “physical state data,” and using a pre-trained health assessment model to output a probability of disease/severity (e.g., COPD). See, e.g., “collect and process… physical state data” and “input… into a health assessment model… probability of suffering from the target disease.” (Spec. [0067], [0030]).
Claims 1-20 recite abstract ideas categorized as mental processes, mathematical concepts, and certain methods of organizing human activity. Under the Broadest Reasonable Interpretation (BRI), independent claims 1 and 17 describe parent functions of information collection and core structures that recite mental steps. These are narrowed by child functions in claims 2-7, such as "pushing a questionnaire" and "binning," which are interpreted as human activity processes and mental evaluations. Child functions in claims 8-15, including "cubic spline interpolation" and "AI models," describe mathematical relationships and calculations. Child structures in claims 18-20 add generic components like "memory" and "storage media" to store and process the abstract data further being analysis under prong two and step 2B.
Recitation of Independent representation Claims; bold are additional elements and non-bold abstract idea
Claim 17.
A data processing device, comprising:
a vital sign data acquisition module, configured to acquire vital sign data of a target object, wherein
the vital sign data includes monitoring data obtained by monitoring the vital signs of the target
object and supplemented data supplemented by the vital sign data according to the monitoring data;
a multimodal data acquisition module, configured to acquire multimodal data of the target object,
wherein the multimodal data includes at least one of image data of the target object and survey data
for preset symptoms; and
a physical state data generating module, configured to generate the physical state data of the target
object according to the vital sign data and the multimodal data.
Claim Abstract Classification Rational (Independent Claims 1 and 17)
Under their Broadest Reasonable Interpretation (MPEP § 2111), the independent claims 1 and 17 recite the abstract idea of collecting disparate physiological and subjective data points to synthesize a comprehensive diagnostic profile of an individual’s health status. This process aligns with the following abstract idea categories:
Mental Process (MPEP § 2106.04(a)(2)(III)): Mental processes are concepts that can be performed in the human mind, including observations, evaluations, and judgments used to reach a conclusion. The independent claims 1 and 17 recite "generating a physical state data of the target object according to the vital sign data and the multimodal data." This step describes an analytical evaluation where a processor (or a clinician) reviews various data inputs to form a judgment regarding a patient's condition. This fits the definition in MPEP 2106 because the "generating" of a state from data is a cognitive exercise in data synthesis that does not require a specialized computer to perform. The specification supports this, stating: "It is possible to obtain more accurate and comprehensive physical data of the target object through limited monitoring means... thereby improving the data collection effect" (Spec., para. [0074]). This paragraph confirms that the "generation" of data is an analytical improvement on data collection efficiency rather than a technical improvement to computer hardware.
Certain Method of Organizing Human Activity (MPEP § 2106.04(a)(2)(II)): This category includes managed workflows and interactions between people, such as social or professional relationships. The independent claims 1 and 17 recite "collecting multimodal data... [including] survey data for preset symptoms." This describes a managed workflow of interaction, which falls under the sub-category of Managing Personal Behavior or Relationships / Interactions. This fits the category because the collection of survey data through questionnaires is a traditional method of patient-provider interaction used to gather subjective evidence. The specification supports this, stating: "The survey data can be set up with corresponding questionnaires for different target diseases... [the] target subject can log in... and fill in the corresponding questionnaire" (Spec., para. [0109-0110]). These paragraphs are relevant as they describe a social/professional interaction (question and answer) that organizes how a human provides information to a system, which is a fundamental method of organizing human activity.
Manual Replication Scenario (Human Equivalence)
The abstract nature of the claims is reinforced because the entire process is analogous to fundamental human activities:
Under MPEP 2106.04(a), a claim is directed to a mental process if it describes a concept that can be performed in the human mind, or by a human using a pen and paper. Although the Applicant’s claims utilize a computer to process data with greater speed and efficiency than a human, the USPTO and the courts have consistently held that the mere automation of an abstract idea does not transform it into a patent-eligible invention. As stated in MPEP 2106.05(a), "the use of a physical computer to perform an abstract idea with more speed and accuracy" does not constitute a technical improvement to the computer itself, but rather an improvement to the abstract process. Therefore, the "generating" of physical state data remains an abstract mental evaluation regardless of the computational power used to execute it.
Step-by-Step Human Analogy for the Independent Claimed Process:
To demonstrate that the independent claims 1 and 17 recite a process that can be performed through manual human activity, consider the following clinical analogy mirroring the limitations of the claims:
Acquiring vital sign data (Claim 1/17 limitation): A nurse monitors a patient's vital signs (e.g., blood oxygen or heart rate) at specific intervals throughout the day and records them in a medical chart. To account for "supplemented data," the nurse notices a missing entry from a 2:00 PM check and uses her medical knowledge to interpolate a likely value based on the 1:00 PM and 3:00 PM readings to ensure the chart is "complete" for the doctor's review.
Collecting multimodal data (Claim 1/17 limitation): A receptionist provides the patient with a COPD screening questionnaire (survey data) and requests the patient's previous CT scan results (image data). The patient fills out the questionnaire—selecting options for symptoms like shortness of breath or smoking history and hands both the survey and the images to the medical staff.
Generating physical state data (Claim 1/17 limitation): A doctor reviews the completed vital sign chart (the monitoring and supplemented data) alongside the patient’s survey answers and CT scan (multimodal data). By mentally synthesizing these disparate data points, the doctor produces a "physical state data" summary—essentially a clinical evaluation of the patient's current health status.
This manual process entirely mirrors the "parent functions" of claims 1 and 17. The specification reinforces this human equivalence by noting that the data collection is intended to help professionals "analyze and predict the user's disease status" (Spec., para. [0003]). This confirms that the claimed invention is merely a computerized version of a standard medical evaluation protocol.
Dependent claims
Claims 2-5: These claims recite under BRI the sub-steps of "binning the monitoring data" and "grouping" based on the "degree of impact... on the target disease" [Claim 3], which is a Mental Process involving the evaluation and judgment of clinical data.
Claim 6: This claim recites under BRI "generating the complementary data... by cubic spline interpolation," which is a Mathematical Concept consisting of mathematical formulas and calculations used to derive new data points.
Claim 7: This claim recites under BRI "pushing a questionnaire" and "receiving survey data" to be used as "external variables" [Claim 7], which is a Certain Method of Organizing Human Activity (Managing Personal Behavior or Relationships / Interactions).
Claim 8Claim 8 is directed to the abstract idea of performing a mathematical evaluation to determine a “probability of suffering from a target disease.” The claim recites inputting data into a health assessment model and outputting a numerical probability, which constitutes a mathematical calculation applied to data. Under MPEP § 2106.04(a)(1), this is a mathematical concept because it merely applies formulas or statistical techniques to generate a quantitative result.
Claim 9Claim 9 recites a health assessment model that applies a sequence of statistical algorithms, including an ARIMA model and an Informer model, to health-related data. These limitations define specific mathematical relationships and time-series forecasting calculations used to transform input data. As such, the claim is directed to a mathematical concept performed by automated algorithms rather than to a technological improvement.
Claim 10Claim 10 further specifies that the first (ARIMA) and second (Informer) models process vital sign data, while the third (N-BEATSx) model processes both vital sign data and multimodal data. This refinement merely assigns different data sets to different mathematical models within the same analytical workflow. The focus remains on organizing and applying mathematical calculations to data, which is an abstract idea under MPEP § 2106.04(a)(1).
Claim 11Claim 11 recites inclusion of a third N-BEATSx model within the health assessment model to further analyze the data. N-BEATSx is itself a mathematical forecasting framework that relies on numerical optimization and statistical modeling. The claim therefore remains directed to a mathematical concept implemented through algorithmic processing.
Claim 12Claim 12 recites a fourth Informer model applied within the same sequence of statistical models. This limitation adds another layer of mathematical analysis but does not alter the nature of the invention beyond additional mathematical calculations. Accordingly, the claim is directed to an abstract idea because it continues to rely solely on mathematical concepts to generate predictive outputs.
Claim 13 recites the abstract idea of mathematical processing of physiological time-series data to derive a predicted trend, by “resampling… into high-frequency data and low-frequency data,” “performing difference processing,” and “inputting… into the fourth model to obtain the prediction result of the periodic trend.” These limitations are mathematical concepts because resampling and differencing are statistical/mathematical transformations applied to data, and the model-based “prediction result” reflects computation of a numerical/trend output from those transformations. Accordingly, under Prong One, Claim 13 is directed to a judicial exception since it centers on data manipulation and mathematical calculations for prediction, rather than a specific technological improvement in the functioning of a computer or medical device.
Claims 14–15 recite the abstract idea of mathematically combining model outputs and training a statistical classifier to generate a disease-probability prediction, by “splicing output results… in time and input into the fifth model for integrated training” and specifying the fifth model is a “LightGBM model.” These limitations are mathematical concepts because they describe data aggregation (splicing) and algorithmic model training/inference that applies statistical/machine-learning calculations to transform inputs into a predictive output. Accordingly, under Prong One, Claims 14–15 are directed to a judicial exception since the focus is on mathematical/statistical computation over data, not on an improvement to computer functionality or another specific technological process.
Claims 18-20: These claims recite under BRI the performance of the "data processing method" steps within a device or storage medium environment. These claims recite Mental Processes and Mathematical Concepts because they incorporate by reference the automated execution of the data collection, interpolation, and algorithmic assessment steps identified in the method claims.
Having identified that the claims as a whole recite judicial exceptions under Step 2A, Prong One, the analysis must now determine whether the additional elements integrate these abstract ideas into a practical application.
Step 2A, Prong Two
Prong Two asks whether the claims, as a whole, add elements that apply the recited abstract idea in a manner that meaningfully limits it to a practical application, rather than merely using the abstract idea in a particular setting.
Evaluation of Independent Claims 1 and 17 Additional Elements
Acquisition and Generating Modules: The recitation of a "vital sign data acquisition module," "multimodal data acquisition module," and "physical state data generating module" fails to integrate the abstract idea into a practical application.
These elements describe generic functional modules used solely to facilitate the mental and mathematical exceptions identified in Prong One. According to MPEP 2106.05(f), integration is not present when the additional elements "do no more than describe the judicial exception... and describe that the judicial exception is applied in a generic computer environment." The specification confirms these modules are merely "division[s] of logical functions" that can be "implemented in the form of software" or "hardware associated with program instructions" (Spec., para. [0180], [0183]). This indicates that the modules serve as a mere conduit for the "analytical and prediction" capabilities of the process, rather than a technical improvement to signal processing or medical sensing hardware.
When viewed as a whole, the combination of these modules reflects an automated version of a clinician’s data organization workflow. The specification acknowledges the goal is to "reduce manual participation and save human resources" by providing results "conveniently and quickly" (Spec., para. [0120]). Under MPEP 2106.05(a), an improvement in the "speed or accuracy" of an abstract process through a generic computer does not integrate the exception into a practical application, as the improvement is to the efficiency of the abstract idea itself rather than to the underlying technology.
Dependent claims analysis
Claims 8–15
The recitation of a "pre-trained model" including "ARIMA," "Informer," "N-BeatXs," and "LightGBM" fails to integrate the abstract idea into a practical application.
These elements fail to improve computer functionality (MPEP 2106.05(a)). While the models provide a refined "health assessment result" [Claim 8], an improvement in the "accuracy of a mathematically calculated statistical prediction" is an improvement to the abstract idea itself, not a technology. The claims do not recite a technical mechanism that reduces computational resource usage or increases processor speed; they merely "use the computer as a tool" to perform complex calculations (MPEP 2106.05(f)). Consequently, these models do not move the claim beyond Step 2A.
Claims 18–20
The recitation of a "memory," "processor," "program," and "readable storage medium" fails to integrate the abstract idea into a practical application.
These are additional elements but reflect "mere instructions to 'apply it' on a computer" (MPEP 2106.05(f)). Under BRI, these components are "invoked to perform the abstract idea" without a specific functional relationship that improves the computer itself. These elements "simply link the expression of the judicial exception to a particular technological environment" [Claims 19, 20], which constitutes "insignificant extra-solution activity" (MPEP 2106.05(h)). There is no specific technical mechanism described that alters how the processor or memory operates or provides a solution to a technical problem in computer technology.
Claims 2–7, 16
The remaining claims do not pass Prong Two as they recite no additional elements beyond the judicial exceptions identified in Prong One.
Limitations such as "binning" [Claim 2], "interpolation" [Claim 6], "pushing a questionnaire" [Claim 7], and narrowing the scope to "chronic obstructive pulmonary disease" [Claim 16] are the judicial exceptions themselves (Mental Processes and Human Activity). Because these limitations define the abstract idea, they cannot serve as the "additional elements" required to integrate that idea into a practical application (MPEP 2106.05). These claims merely narrow the abstract process to a specific medical field.
When viewed as a whole, the combination of these elements does not pass Prong Two. The claims link generic hardware with mathematical refinements to notify a user of a disease probability, which is an improvement to a medical conclusion rather than a technological process. Because Step 2A is not satisfied, the analysis must proceed to Step 2B.
Step 2B: Inventive Concept Analysis (MPEP §2106)
Step 2B asks whether the additional elements (identified in Prong Two) “amount to significantly more than the judicial exception” when evaluated individually and in combination, after construing the claim under BRI.
Step 2B determines whether the additional elements, alone or in combination, amount to an "inventive concept" that is "significantly more" than the judicial exception. The claims fail this requirement because the recited components are invoked solely to automate a medical evaluation process without providing a specific technical solution to a computer-centric problem.
Evaluation of Independent Claims 1 and 17 Additional Elements
Functional “Modules” / Generic Implementation: The recitation of a “vital sign data acquisition module,” “multimodal data acquisition module,” and “physical state data generating module” fails to provide an inventive concept.
These elements represent "mere instructions to 'apply' the exception" on a computer (MPEP 2106.05(f)). The specification explicitly admits these are not specialized hardware but rather “a division of logical functions” that can be “fully or partially integrated into one physical entity” or “implemented in the form of software” (Spec., para. [0183]). Consequently, these modules do not reflect a technological mechanism; they are functional placeholders for the "analytical and prediction" capabilities of the clinician’s mental process, which is insufficient to supply "significantly more" (MPEP 2106.05(g)).
When viewed as a whole (Claims 1 and 17): The combination of these modules reflects a generic automation of collecting health information to reach a diagnostic conclusion. The specification confirms the goal is to “obtain more accurate and comprehensive physical data... thereby improving the data collection effect” (Spec., para. [0074]). Under MPEP 2106.05(a), an improvement to the efficiency of the abstract idea itself (the medical evaluation) does not constitute an inventive concept for the underlying computer technology.
Dependent Claims Analysis
(Claims 8–15): “Pre-trained model” / ARIMA / Informer / N-BeatXs / LightGBM
These claims add specific algorithmic labels and training steps which fail to provide an inventive concept.
These limitations represent "insignificant pre-solution activity" (MPEP 2106.05(g)). The specification describes using these models—such as the “ARIMA model,” “Informer model,” and “LightGBM model”—to “accurately and quickly generate preliminary prediction results” (Spec., para. [0121], [0135]). Per Stanford II, an improvement in the "accuracy of a mathematically calculated statistical prediction" is an improvement to the abstract idea itself, not a technological improvement. Because the claims do not recite a technical mechanism that alters the computer's internal functioning (e.g., reducing memory footprint or increasing processing speed), they remain mere instructions to apply math on a computer.
“Memory,” “Processor,” “Program,” “Readable Storage Medium” (Claims 18-20)
These claims add generic computer hardware that fails to provide an inventive concept.
These elements reflect a "mere field-of-use limitation" (MPEP 2106.05(h)). The specification confirms the hardware comprises a “processor... memory, and a program” where the processor is “configured to read the program in the memory” (Spec., para. [0051]). These components perform their "well-understood" functions to execute the abstract method identified in Prong One. There is no functional relationship described that improves the hardware itself; rather, the hardware is “associated with program instructions” to perform the process (Spec., para. [0180]).
Claims without new additional elements (Claims 2–7 and 16):
These claims recite no additional elements beyond the judicial exceptions identified in Prong One (e.g., "binning," "cubic spline interpolation," or "questionnaires"). As these limitations define the abstract idea, they cannot contribute an inventive concept (MPEP 2106.05). They merely narrow the “medical or clinical reference significance” of the data within the specific field of “chronic obstructive pulmonary disease” (Spec., para. [0085], [0109]).
Considering the additional elements together (functional modules + trained models + generic hardware), the claims amount to a computerized version of a standard clinical data-organization workflow. The combination lacks an inventive concept because the elements perform their routine functions to facilitate the automated execution of mental processes and mathematical calculations.
The claims are directed to judicial exceptions and, considering the additional elements individually and in combination, lack an inventive concept under Step 2B. Therefore, Claims 1–20 are rejected under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-3, 8 and 17-20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated over Shulzas - US20220301666A1.
Shluzas teaches Claim 1.
A data processing method, comprising: (Shluzas, abstract, par. 0004)
Shluzas et al. shows a method that takes in physiological and other patient information and processes it to produce a result, which is a data processing method.
acquiring vital sign data of a target object, wherein the vital sign data comprises monitoring data obtained by monitoring vital signs of the target object and supplemented data obtained by supplementing the vital sign data according to the monitoring data; (Shluzas, par. Par. 0056 “processing patient data ( both time series data and crosssectional data, … by patient’s device, par. 0022 ” time - series physiological data is captured in real - time by sensors worn by the patient”, par. 0168, “Data Imputation : Missing patient measurements can be input with previous day measurement for the same patient if available …”)
Shluzas et al. shows monitoring-based acquisition because it expressly captures patient physiological measurements from sensors as the underlying monitoring data, and it shows supplementation because it expressly fills missing measurements by inputting values derived from the patient measurement history to compensate for missingness and irregular sampling
collecting multimodal data of the target object, wherein the multimodal data comprises at least one of image data of the target object and survey data for preset symptoms; (Shluzas, fig. 26, par. 0089, “Cross - sectional data may include image data “, par. 0119, 0123, 0197 “inputs ( speech , images ) from a smartphone or other wear able devices” )
Shluzas et al. read on, collecting multimodal data for the target object where the multimodal data includes at least image, audio and text data of the target object
and generating physical state data of the target object according to the vital sign data and the multimodal data. (Shluzas, par. 0013, “outputting a patient score “)
Shluzas et al. shows the claimed generating because it expressly uses time series physiological data as the vital sign data and uses cross sectional data that includes image data, then processes those inputs in a trained machine learning model to output a patient score that assesses the patient’s health, which is physical state data generated according to both inputs,
Shluzas teaches, Claim 2.
The method according to claim 1, wherein the obtaining the vital sign data of the target
object comprises:
acquiring monitoring data obtained by monitoring the vital signs of the target object; (Shluzas, fig. 26, par. 0089, “Cross - sectional data may include image data “, par. 0119, 0123, 0197 “inputs (speech , images) from a smartphone or other wear able devices”, par. 0072 “recording physiological data from the patient”, par. 0144 “capture vital signs” )
binning the monitoring data; (Shluzas, par. 0157 “data processing, cleaning, stratifying, and/or prioritization”, par. 0168)
Shluzas et al. describes an AI module that performs stratifying.
grouping the monitoring data after binning according to a preset monitoring period; (See at least, Paragraph 0170, the time series data were sorted by date and time and Paragraph 0180, outputs a daily prediction score., [0174-0176])
Shluzas et al. describes daily prediction scores and time-sorted sequences and dividing data into window intervals (i.e. grouping the data based on a present monitoring period)
generating supplementary data to supplement the missing monitoring data in each monitoring cycle; (Shluzas, par. 0168, “Missing patient measurements can be input with previous day measurement” -0169 “Missing measurements may be input to maintain a constant sampling interval to compensate for irregular sampling” )
Shluzas et al. describes how missing measurements are replaced with previous day measurements or median values.
and
using the monitoring data and the supplemented data as the vital sign data of the target object. (Shluzas, par. 0168-0171, 0180)
Shluzas et al. discloses the step where the system stops treating observed measurements and gap-filled values as separate tracks and instead uses them together as one operative patient dataset for model input. Shluzas et al describes vital sign measurements that exist in the raw dataset, then fills gaps by inputting missing patient measurements using prior measurements or medians, and then keeps those imputed values in the time-series so the model receives a continuous vital sign sequence, so the resulting vital sign data necessarily includes both monitored measurements and supplemented imputed values as one input stream for analysis and scoring
Shluzas teaches, Claim 3.
The method according to claim 2, wherein the binning the monitoring data comprises:
binning the monitoring data according to a degree of impact of the monitoring data on the target disease. (Shluzas par. 0157, 0187, 0008, abstract)
Shluzas describes a system where health data is sorted using a color-coded hierarchy. Each color represents a specific level of medical urgency or injury severity. This color-coding functions as a bin. Because the system assigns these colors based on the importance of the medical data to the survival or health of the patient, it is sorting the monitoring data into groups according to its impact on the injury or disease state. This process of assigning a category like red or yellow based on injury severity performs the exact function of binning data by its degree of impact.
Shluzas teaches, Claim 8.
A health assessment method, comprising:
acquiring physical state data of the target object, wherein the physical state data is obtained by the data processing method according to claim 1; Refer to Claim 1 rejected under 35 U.S.C 102.
inputting the physical condition data into a health assessment model to obtain a health assessment result of the target object suffering from a target disease, wherein the health assessment model is a pre-trained model that takes the physical condition data as input and the probability of suffering from the target disease as output. (Shluzas, par. 0174, 0178, 0230-0231)
Shluzas et al. describes inputting patient data into a trained machine learning model and using the model’s output probability to determine whether the patient is positive for a condition which reads on obtaining a health assessment result with probability output for a target disease.
Note: Claims 17, 19-20 are rejected with claim 1 and claim 18 with claim 8 for being very similar.
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 text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claim 4 are rejected under 35 U.S.C. 103 as being unpatentable over Shulzas - US20220301666A1 and further in view of Narziev, N., Goh, H., Toshnazarov, K., Lee, S. A., Chung, K.-M., & Noh, Y. (2020). STDD: Short-term depression detection with passive sensing. Sensors, 20(5), 1396, refer to PTO-892 U, in further view of Lustgarten, J. L., Visweswaran, S., Gopalakrishnan, V., & Cooper, G. F. (2011). Application of an efficient Bayesian discretization method to biomedical data. BMC Bioinformatics, 12, 309. Refer to PTO-892-U.
Shluzas teaches, Claim 4.
The method according to claim 3, wherein the binning the monitoring data comprises:
binning the monitoring data using the .
Shluzas teaches the general method of monitoring health data and "binning" said data through stratification and windowed intervals to establish a physiological baseline (Shluzas [0157], [0176]). Shluzas further teaches optimizing the AI model using entropy-based criteria (Binary Cross-Entropy) ([0231]).
Shluzas does not explicitly disclose that the binning method is a "minimum entropy binning method."
Lustgarten teaches the Minimum Entropy Binning Method, that required a specific supervised discretization algorithm that minimizes class entropy. Lustgarten discloses that the Fayyad and Irani (FI) method is a "widely used, abstract" discretization technique that "minimizes the joint entropy of the two resulting subintervals, page 3" to determine the optimal cut points. Lustgarten further defines this as finding "the discretization with the minimum entropy, page 3" and notes it is a "standard algorithmic benchmark, page 5" for processing biomedical data.
The combination of Shluzas + Lustgarten makes obvious the full limitation binning the monitoring data using the minimum entropy binning method because a POSITA would implement the minimum entropy binning method (FI) of Lustgarten as a routine data preprocessing configuration within Shluzas's health monitoring workflow. Shluzas seeks to "establish a patient's physiological data baseline, par. 0157" from continuous sensor data, and Lustgarten teaches that Minimum Entropy discretization is a "standard algorithmic benchmark, page 5" that "increases the accuracy of classifiers, page 10" on such biomedical data.
A skilled Artisan in the art who read Shluzas application, would combine Lustgarten with Shluzas, because Shluzas encounters the problem of processing raw continuous sensor data to determining a patient's state, where the solution is found in the secondary prior art Lustgarten which teaches that applying "minimum entropy" discretization specifically improves the performance of classifiers on such biomedical data. Shluzas operates in the same field of analyzing physiological signals but relies on general stratification, whereas Lustgarten provides the specific "standard algorithmic benchmark" to optimize those intervals. (Reference, See at least Shluzas, Paragraph [0157] "establishing a patient's physiological data baseline... comprising collecting, by the sensor unit, the physiological data"; Lustgarten, Abstract "Several data mining methods require data that are discrete... [FI method] is commonly used for discretization"; Page 2, "discretization has been shown to increase the accuracy of some classifiers").
A skilled artisan would have a Reasonable Expectation of Success because Lustgarten explicitly validates that this specific binning method is a "standard" and "widely used" technique that yields statistically significant improvements on the exact type of "biomedical datasets" used by Shluzas. It is not an unproven experiment; it is a routine substitution of a superior discretization algorithm.
Claim 5 and 7 are rejected under 35 U.S.C. 103 as being unpatentable over Shulzas - US20220301666A1 and further in view of Narziev, N., Goh, H., Toshnazarov, K., Lee, S. A., Chung, K.-M., & Noh, Y. (2020). STDD: Short-term depression detection with passive sensing. Sensors, 20(5), 1396, refer to PTO-892 U
Shluzas teaches, Claim 5.
The method according to claim 2, wherein after the
monitoring data after binning is grouped according to a preset monitoring period, the method further comprises:
detecting a ; (Shluzas, par. 0168-0169, 0157, 0175)
Shluzas teaches, of the Claim 5, wherein after the monitoring data after binning is grouped according to a preset monitoring period, the method further comprises: detecting… first cycles in the monitoring cycle, wherein the first cycle is a monitoring cycle in which monitoring data at a target time is missing, that required identifying specific time blocks within a dataset where an expected data point is absent at a scheduled moment. Shluzas discloses monitoring data at a target time is missing because it identifies when specific measurements are absent within a defined interval to trigger data cleaning. Shluzas manages time-series data where expected values are absent at specific sampling points, which constitutes identifying a missing condition at a target time within a period. (Shluzas, See at least, para. [0169]: “Missing measurements may be input to maintain a constant sampling interval to compensate for irregular sampling.”). However, Shluzas does not describe detecting a number of first cycles, because while it identifies missing measurements for compensation, it does not explicitly perform the mathematical act of tallying the total count of those cycles.
Narziev, read on broadest reasonable interpretation detecting a number of first cycles that required generating a quantitative count of intervals that failed to provide data at a scheduled check-in. Narziev quantifies missingness by tracking “skips” at scheduled times and reporting the resulting total count of valid vs. missing periods. (Narziev, See at least, Page 10: “data loss” and Page 11: “response rate table #3.”).
A skilled Artisan in the art who read Shluzas application, would combine Narziev with Shluzas, as both are in the same subject matter of remote health monitoring and Shluzas identifies “irregular sampling, par. 0129” as a problem, where the solution of quantifying those irregularities is found in Narziev. (Reference, See at least, Shluzas para. [0169] and Narziev Page 8: “data provision performance., page 8”). This integration resolves the problem of assessing data reliability by providing a count of missed cycles, which serves as a benefit to the clinician to determine if the “predictive analytics” are based on sufficient data. (Reference, See at least, Shluzas para. [0105]). The artisan would have a Reasonable Expectation of Success because counting the occurrences of a condition already recognized by Shluzas (missing measurements) using Narziev’s accounting method is a routine computational task that yields the predictable result of a data-integrity tally.
Shluzas teaches, Claim 7.
The method according to claim 1, wherein the collecting multimodal data of the target
object comprises:
(Shulas, par. 0131, “ The caregiver would enter clinical data via voice , touch , or button - based input “)
(Shluzas, par. 0128, “selecting an item from a drop - down menu or dragging and dropping icons “)
Shluzas demonstrates standard user interface actions like using menus to navigate or select items.
standardizing the survey data according to preset rules to form the multimodal data, wherein the standardized(Shluzas, par. 0204, “ reducing the initial dimensionality of the raw data down “, par. 0168 “Data Imputation: Missing patient measurements can be input with previous day measurement for the same patient if available. Otherwise, the median measurement for that patient across all available measurement dates can be used”, par. 0081 “background information, physiological data, or patient history data that may be useful in classifying, diagnosing and/or treating a patient's medical condition” )
Shluzas teaches receiving survey data input and selection input of at least one option because it describes a user interface where clinical data is captured through specific manual interactions. Shluzas describes using dropdown menus and touch inputs to record patient states, which functions as capturing selections from multiple options to document health. (Shluzas, See at least, the caregiver would enter clinical data via voice touch or button-based input 0131, selecting an item from a drop-down menu or dragging and dropping icons 0128, reducing the initial dimensionality of the raw data down 0204).
However Shluzas, does not describe pushing a questionnaire targeting the preset symptoms to the target object or standardizing the survey data... used as external variables.
Narziev, teaches, pushing a questionnaire and standardizing the survey data... used as external variables that required a triggered delivery of queries and the conversion of responses into fixed model features. Because Narziev describes delivering questionnaires via smartphone notifications and using the resulting ordinal selections as dataset features. This teaches the automated "push" of a survey and the standardization of those inputs into a vector format used by a classification model. (Narziev, See at least, EMA questionnaires were delivered according to the schedule six times per day through notifications to the participants smartphones Section 3.4, page 6, For every question they were provided a scale of ordinals from 1 to 10 Section 4, page 7, we utilized all the features... along with EMA responses from participants to build a dataset for training Section 4.1.2 page 10).
A skilled Artisan in the art who read Shluzas application, would combine Narziev with Shluzas because they are in the same subject matter of remote health monitoring (Shluzas, par. 0012 and Narziev page 2) and Shluzas identifies a problem with incomplete documentation in complex clinical scenarios, where a solution of capturing self-reported symptom clusters via "push" notifications is found in Narziev.
The integration of Narziev's pushed questionnaires resolves the problem of missing subjective symptom data in Shluzas, providing a benefit of a more robust multimodal dataset for health assessment with expected predictive results. They would have a Reasonable Expectation of Success because both systems utilize smartphone-based interfaces, use machine learning models that accept feature vectors, and process structured ordinal data, making the software integration a predictable application of existing technologies.
Claim(s) 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Shulzas - US20220301666A1 and further in view of Narziev, N., Goh, H., Toshnazarov, K., Lee, S. A., Chung, K.-M., & Noh, Y. (2020). STDD: Short-term depression detection with passive sensing. Sensors, 20(5), 1396, refer to PTO-892 U, in further view of Câmara, A. J. A. (2019). Modelo aditivo generalizado para dados de contagem: Uma aplicação para avaliar o impacto da poluição atmosférica na saúde [Master's thesis, Universidade Federal de Minas Gerais]. Refer to PTO-892 V
Shluzas teaches, Claim 6.
The method according to claim 2, wherein the generating the supplementary data for supplementing the missing monitoring data in each monitoring cycle comprises:
generating the complementary data of the monitoring data by . (Shluzas, see at least par. 0174-0175, assigned dates using a standard interpolation method and missing patient measurements can be input with previous day measurement... or the median..)
Shluzas shows how to handle missing health records by using the median of a group or applying a basic interpolation method for dates. While this fills in missing values, Shluzas does not specify using the particular cubic spline curve-fitting math required to ensure the generated data segments connect smoothly with continuous derivatives.
Câmara teaches by cubic spline interpolation of the Claim 6, missing from Shluzas that required a mathematical curve-fitting technique that uses third-order polynomials to connect data points in a way that ensures the resulting curve is smooth and continuous. Camara teaches spline interpolation using B-splines, and explains the cubic spline as the common spline used because it is “fairly smooth” and enforces “continuity restrictions up to the second derivative.” (Camara, The most common spline is a cubic spline …cubic spline on [a, b], if s is a cubic polynomial si in each interval [wt , wt+1], page 8..)
A skilled Artisan in the art who read the Shluzas application would combine Câmara with Shluzas because they are in the same subject matter of physiological data processing and Shluzas presents a problem (gaps in patient measurements), where a solution is found in the secondary prior art Câmara which provides a refined mathematical method for data continuity. Specifically, Shluzas recognizes the need to handle missing data to ensure a complete health record, and Câmara provides the specific "cubic spline" technique used to model health-related time-series data. (Reference Shluzas, See at least, par. [0174] “patient measurements can be input... using a standard interpolation method”;
Because Shluzas seeks the goal of filling gaps in patient data to ensure a complete health record and Câmara teaches the cubic spline technique that predictably achieves that goal on the same type of health-related time-series input. Following MPEP 2143 (Rationale B), substituting the generic "standard interpolation" of Shluzas with the specific "cubic spline" of Câmara yields the predictable result of a mathematically smooth and continuous data transition. (Reference Shluzas, See at least, par. [0174] “patient measurements can be input... using a standard interpolation method”; Reference Câmara, See at least, Page 46 where the code creates the smooth curve: “as.data.frame(ns (DataRD_filter+Tmin, df=3))” and the title “APLICAÇÃO PARA AVALIAR O IMPACTO... NA SAÚDE” [using this smooth model to evaluate health impact]).
A POSITA would reasonably expect success because cubic spline interpolation is not a speculative modification. It is a known, routine interpolation method for numeric sequences, built from polynomial pieces “joined” at knots, and explicitly described as the common spline due to smoothness constraints.
Shluzas in combination with Narziev teaches, Claim 16.
The method according to claim 8, wherein the target disease is . (Shluzas, par. 0219-0220) Shluzas describes detecting and assessing diseases such as the flu or with COVID - 19 and lists additional disease application areas such as sepsis / septic shock detection and mortality risk , acute respiratory failure and mortality risk. Shluzas does not state that the target disease is chronic obstructive pulmonary disease, and Shluzas does not describe a COPD specific detection or COPD probability output.
Camara teaches evaluating the impact of environmental factors on Chronic Obstructive Pulmonary Disease (COPD) of the Claim 16, wherein the target disease is chronic obstructive pulmonary disease that required the selection of COPD as the specific respiratory condition for analysis. Camara discloses statistical models to "evaluate the impact of air pollutants and meteorological variables on the number of Chronic obstructive pulmonary disease (COPD) cases, page 5". Camara teaches that "Epidemiological data are frequently treated as time series... for modelling this type of data" and explicitly links "particulate matter (PM)" and "pollutant concentration levels" to "respiratory... diseases". (Reference Camara, See at least, Abstract, Introduction page 2, Conclusion, page 32)
The combination of Shluzas + Camara applications make obvious the full limitation wherein the target disease is chronic obstructive pulmonary disease because assessing COPD is a simple substitution of one known respiratory condition for another within the same predictive framework. A POSITA would implement COPD detection (as taught by Camara) as a routine configuration to make the disease prediction model operate within Shluzas's patient monitoring system, because Shluzas seeks to predict health outcomes based on "lung particulate exposures" and "environmental/military occupational exposure" and Camara teaches that these same inputs (air pollutants, particulate matter) are directly correlated with COPD cases that predictably achieves the goal of expanding the system's diagnostic capabilities to include prevalent respiratory conditions caused by the very exposures Shluzas monitors. (Reference, See at least, Shluzas par. [0083], [0093]; Camara Abstract, Conclusion).
A skilled Artisan in the art who read Shluzas application, would combine Camara with Shluzas, as the primary art suggested monitoring "pulmonary exposures limiting performance, par. 0083" and "lung particulate exposures, par. 0093”, where a solution are in the secondary prior art which provides the specific correlation between these particulate exposures and COPD. (Reference, See at least, Shluzas par. [0083]; Camara Conclusion).
Reasonable Expectation of Success because Shluzas already employs "RNN, LSTM, GRU" models for "time series data" and Camara demonstrates the successful statistical modelling of COPD using similar time-series environmental data, making the adaptation of the target disease variable a matter of routine data training.
Claim(s) 9-15 are rejected under 35 U.S.C. 103 as being unpatentable over Shulzas - US20220301666A1 and further in view of Narziev, N., Goh, H., Toshnazarov, K., Lee, S. A., Chung, K.-M., & Noh, Y. (2020). STDD: Short-term depression detection with passive sensing. Sensors, 20(5), 1396 Refer to PTO-892-U and further view of Olivares, K. G. (2023). Applied mathematics of the future Doctoral dissertation, Carnegie Mellon University PTO-892-W
Shluzas teaches, Claim 9.
The method according to claim 8, wherein the health assessment model comprises an
integrated first model, a second model and a third model, wherein the first model is an . (Shluzas, See at least, Ensemble approaches have yielded favorable results... by aggregating the predictions from multiple weak learners/different models 0227, the RNN ensemble method utilizes a voting system that heuristically chooses the "most different" models 0228).
model, the second model is an . (Shulazas, par. 0227- 0228)
Shluzas teaches the use of an integrated first model, a second model and a third model because it describes using ensemble approaches that combine multiple different models into a single predictive output. Shluzas explains that aggregating predictions from various "weak learners" or "different models" improves accuracy and sensitivity, which describes the structure of an integrated multi-model assessment
However Shluzas, does not describe the specific selection where the first model is an ARIMA model, the second model is an Informer model, and the third model is an N-BeatXs model. Shluzas provides the general framework for an ensemble of "different models" (such as LSTMs and GRUs) but does not explicitly name these three specific architectures as the components of the ensemble.
ARIMA: Oliver teaches ARIMA of the Claim 9, as basic time series foresting models, use to evaluate the predictive accuracy and to autocorrelations present in the data.
(Refer to pag. 6, 1.6 section HierarchicalForecast… A benchmark library for hierarchical
forecasting …to improve the availability, utility, and adoption of hierarchical forecast reference baselines. Page. 9, section 2.2.1 “to evaluate the predictive accuracy”, page. 21, section. 3.1.4 “focus on characterizing the autocorrelations present in the data.” )
A skilled Artisan in the art who read the Shluzas application would combine Olivares with Shluzas, because they are in the same field of endeavor regarding predictive modeling for time-series data, and Shluzas expressly suggests the strategy of aggregating different models to solve the problem of prediction sensitivity.
The combination of Shluzas + oliver applications make obvious the full limitation the health assessment model comprises an integrated first model, a second model and a third model, wherein the first model is an ARIMA because a POSITA would implement ARIMA as a routine configuration to make the ARIMA modeling of Oliver operate within the integrated ensemble workflow of Shluzas. Because Shluzas seeks the goal of reliability by “aggregating the predictions from multiple... different models, par. 0227” and Oliver teaches the ARIMA technique that predictably achieves a " Oliver, pag.21; the target variable with a linear combination of its past" way to forecast time-series health data. Following, combining these known elements yields the predictable result of a diverse predictive system where the established ARIMA baseline from Oliver to evaluate the predictive accuracy improves the robustness of the integrated health assessment.
Reasonable expectation of success is present because the modification is a straightforward model-selection choice inside an already-disclosed ensemble workflow.
Informer and N-Beatxs:
Olivares teaches the second model is an Informer model, and the third model is an N-BeatXs model of Claim #9, specifically teaching the missing model types required by the claim language.
Informer Model:
Olivares explicitly describes the Informer model within the context of long-horizon forecasting architectures because describe "Informer ... [is a] Transformer with MLP based multi-step prediction strategy, that approximates self-attention with sparsity." (Olivares, See at least, p. 76, par. 6.3). That reads on the claim element "second model is an Informer model" by identifying the specific neural architecture (Transformer-based with sparse self-attention) used for the predictive task.
N-BeatXs Model: Olivares teaches the N-BeatXs (NBEATSx) model as a specific improvement that integrates exogenous variables into a basis expansion analysis. Because describe "We introduced NBEATSx, a neural forecasting solution that extends the neural basis expansion analysis incorporating exogenous variables... The NBEATSx framework decomposes the objective signal by performing separate local nonlinear projections of the data onto basis functions across its different blocks." (Olivares, See at least, page 2, page 36 ). That reads on the claim element "third model is an N-BeatXs model" by disclosing the specific algorithmic structure (basis expansion with exogenous inputs) required.
Integration/Ensembling: Olivares teaches the motivation and method for integrating these distinct models into a single system. Because describe "combining a diverse group of models can be a powerful form of regularization to reduce the variance of predictions... [using] arithmetic mean as the aggregation function." (Olivares, See at least, p. 48, par. 4.5.5). Because this teaches the "integrated" aspect of the claim, showing that a POSITA would combine these specific "diverse" models (Informer and NBEATSx) using an aggregation function to improve predictive performance.
A skilled Artisan in the art who read the Shluzas application would combine Olivares with Shluzas, because they are in the same field of endeavor regarding predictive modeling for time-series data, and Shluzas expressly suggests the strategy of aggregating different models to solve the problem of prediction sensitivity.
The combination of Shluzas + Olivares applications make obvious the full limitation “model, the second model is an Informer model, and the third model is an N- BeatXs model” because a POSITA would implement the specific Informer and N-BeatXs architectures taught by Olivares as the specific instances of the "different models" or "weak learners" utilized in the Shluzas ensemble framework. This modification represents a simple substitution of one known element for another to obtain predictable results. A POSITA would implement the Informer and N-BeatXs models as a routine configuration to make the Olivares advanced forecasting techniques operate within the Shluzas patient monitoring workflow, because Shluzas explicitly seeks the goal of improving predictive sensitivity through aggregation, and Olivares teaches that these specific models are state-of-the-art architectures that predictably achieve that goal on the same type of time-series input. (Reference Shluzas + Olivares, See at least, Shluzas: “aggregating predictions from various ‘weak learners’ or ‘different models’ improves accuracy and sensitivity” [0228]; Olivares: “the proposed ensemble... integrates... Informer... and NHITS page 122, [N-Beats]... significantly improves accuracy in long-horizon forecasting tasks, page 73”; “combining a diverse group of models can be a powerful form of regularization” [4.5.5], page 48).
A skilled artisan would be motivated to combine these references with a Reasonable Expectation of Success because Olivares demonstrates that these specific architectures achieve “state-of-the-art (SOTA), page 48” performance on standard benchmarks, confirming that the combination is not merely theoretical but a provide an enhancement for time-series forecasting. (Olivares, p. 20, Para. 1.2; p. 83, Para. 6.5.4).
Shluzas in combination with Narziev + Olivares teaches, Claim 10.
The method according to claim 9, wherein the input data of the first model and the
second model comprise the vital sign data; (Shluzas, See at least RNN voting ensemble model with ADD aggregation that inputs cross section and time series data, consisting of static and dynamic features, [0017], [0228], FIG. 30.)
Shluzas describes using more than one model by teaching a voting ensemble with multiple voting models and aggregation of their predictions. In that architecture, the first model and the second model are met by at least two of the voting models because Shluzas expressly requires voting models whose outputs are aggregated. Shluzas also describes that the model input includes vital sign data because it lists systolic blood pressure, diastolic blood pressure, temperature, heart rate, and SpO2 as model features.
the input data of the third model comprises the vital sign data and the multimodal data.( Shluzas, See at least Artificial intelligence Machine Learning model may be singular or may employ multi-modal processing where multiple data sources (time-series and cross-sectional) are combined, [0103], [1021], fig. 30-31, 0017)
Shluzas describes using a third model within its architecture by teaching a voting ensemble where multiple data sources (time-series and cross-sectional) are combined to provide a more accurate model of actual patient status. Shluzas explicitly defines this as multi-modal processing. Because the ensemble is composed of multiple trained neural network classifiers (reading on first, second, and third models), and Shluzas teaches that these models use a combination of heart rate, blood pressure, and temperature (reading on vital sign data) alongside cross-sectional data (reading on multimodal data), the limitation is met. Figures 30 and 31 further confirm that the output probability from these combined data streams is fed into the ensemble's constituent models.
Shluzas in combination with Narziev + Olivares teaches, Claim 11.
The method according to claim 9, wherein the health assessment model further
comprises a fourth model, wherein the fourth model is a model which takes the resampled vital sign data as input and takes the probability of suffering from the target disease as output. (Shluzas, See at least, ensemble approaches have yielded favorable results by aggregating the predictions from multiple weak learners/different models 0227, patients who may be missing symptom onset/diagnosis/recovery dates can be assigned dates using a standard interpolation method 0175, artificial intelligence machine learning model is used in combination with predictive analytics to assess a patient’s condition ,Abstract.)
Shluzas describes an ensemble system composed of many different models or weak learners that each contribute a prediction. This plurality of models encompasses the fourth model. The system cleans physiological data by using interpolation to fill gaps in time-series measurements, which constitutes resampling. These adjusted measurements are then processed by the models to generate a predictive assessment or health score. Because a health score or prediction regarding a condition like sepsis signifies the chance of the disease being present, Shluzas describes a model outputting a probability based on resampled health inputs.
Shluzas in combination with Narziev + Olivares teaches, Claim 12.
The method according to claim 11, wherein the fourth model is an Informer model.( Shluzas, See at least, the trained machine learning model may include a neural network a decision tree a random forest 0187, the RNN ensemble method utilizes a voting system that heuristically chooses the most different models 0228. Olivares, See at least, Informer... Transformer with MLP based multi-step prediction strategy, that approximates self-attention with sparsity... The proposed ensemble... integrates... Informer... Section 6.3, refer claim 9 rationale)
Shluzas in combination with Narziev + Olivares teaches, Claim 13.
The method according to claim 11, wherein the obtaining a health
assessment result of the target subject suffering from the target disease by using the health
assessment model comprises:
resampling the vital sign data into high-frequency data and low-frequency data, wherein a sampling frequency of the high-frequency data is greater than the sampling frequency of the low-frequency data, and a sampling frequency of the low-frequency data is no less than twice in each monitoring cycle;( Shluzas et al., See at least, par. 0176, “heart rate measurements were captured at irregularly - spaced intervals but typically once every 15 seconds and the data points can be down - sampled within the windowed interval by taking the median measurement for each day in the sequence and par. 0169 “Missing measurements may be input to maintain a constant sampling interval to compensate for irregular sampling”, par. 0181-0182, 0229)
Shluzas records heart rate data every 15 seconds, which constitutes high-frequency data. The system then processes this same data by down-sampling it to find a daily median or an hourly average. Because 15 seconds is more frequent than hourly or daily rates, Shluzas generates both high-frequency and low-frequency data from the original measurements. By calculating moving averages for 2 or 4 hour windows and providing hourly averaged inputs for the model, the system ensures that multiple data points such as 24 hourly points in a daily cycle are available.
performing difference processing on the low-frequency data according to the sampling frequency of the high-frequency data;( Shluzas, See at least, missing measurements may be input to maintain a constant sampling interval to compensate for irregular sampling 0169, heart rate measurements were captured at irregularly-spaced intervals but typically once every 15 seconds 0176, missing symptom onset/diagnosis/recovery dates can be assigned dates using a standard interpolation method 0175.)
Above limitation require mathematical synchronization of low-frequency or irregular data streams into a high-frequency time-series grid. Shluzas defines the faster-collected data as physiological measurements like heart rate recorded at 15-second intervals. It defines the slower-collected data as irregular clinical milestones like symptom onset, diagnosis, or recovery dates which occur at much slower or inconsistent rates. The system performs the adjustment by inputting missing measurements to maintain a constant sampling interval. By applying interpolation to assign specific dates to these clinical milestones, Shluzas snaps the slower data points onto the high-speed 15-second physiological grid. This process creates a uniform chronological grid where disparate data flows are temporally aligned.
and inputting the high-frequency data and the low-frequency data after difference processing
(Shluzas, See at least, inputting the cross-sectional data and the time-series physiological data into a trained machine learning model 0013, missing measurements may be input to maintain a constant sampling interval to compensate for irregular sampling 0169, patients missing symptom onset diagnosis recovery dates can be assigned dates using a standard interpolation method 0175, the static and dynamic vectors are concatenated 05 and passed as input into fully-connected classification layers 06 to obtain the model prediction for the current time step 0221, he machine learning model is a recurrent neural network (RNN) voting ensemble model that takes a combination of static and dynamic features as input, with daily time steps for the vital sign sequences, and outputs a daily prediction score... par. 0180 .)
Shluzas describes a system where high-speed physiological data and low-speed clinical data are merged for analysis. The high-speed data flows from sensors while the low-speed data consists of irregular events like a diagnosis or medical history. Before these data types are used together, the system performs an adjustment by filling in gaps or using interpolation to align the irregular events with the high-speed timeline. Once this processing creates a synchronized data set, both the original high-frequency stream and the adjusted low-frequency stream are fed into a machine learning model to calculate a patient health score.
respectively into the fourth model to obtain the prediction result of the periodic trend of the vital sign data. (Shluzas, See at least, machine learning algorithm for generating health risk predictions that uses a recurrent neural network RNN voting ensemble model with ADD aggregation 0060, for mortality risk the smartphone application provides a mortality risk prediction score as a function of time in days 0178, outputs a daily prediction score for mortality risk 0180.)
Shluzas teaches a recurrent neural network model that processes data to output a score representing patient health status across a timeline
Shluzas in combination with Narziev + Olivares teaches, Claim 14.
The method according to claim 11, wherein inputting the physical state data into a health assessment model to obtain a health assessment result of the target subject suffering from a target disease comprises:
splicing output results of the first model, the second model, the third model and the third model in time and input into the Shluzas, par. 0221, 0227, 0229, 0206,-0207)
Shluzas shows combining predictions from multiple models using an ensemble voting approach, shown by aggregating the predictions from multiple weak learners / different mod els . and utilizes a voting system.
Shluzas does not show splicing model outputs in time and then using those spliced outputs as the training input to a separate fifth model, because Shluzas describes aggregation among voting models, not a separate fifth model trained on time spliced outputs from first, second, third, and fourth models for example disclosed Speech (ASR), Vision (AOD), Time-Series (RNN), and Static Data (Linear Embedding) models.
Olivares teaches splicing output results... in time and input into the fifth model for integrated training of Claim 14, splicing output results of the first model, the second model, the third model and the third model in time and input into the fifth model for integrated training, that required arranging the predictions from multiple base models chronologically and using that sequence as the training data for a final meta-learner. Olivares describes a meta-learning framework where different models (like ARIMA, Informer, or N-Beats) act as "base learners" whose outputs are integrated to train a superior global model (Olivares, See at least, [1.1], pag. 1 "A global model is fitted simultaneously... sharing information across them... translates in more accurate forecasts"; [1.3], page 2 "NBEATSx improves accuracy through the integration of multiple information sources"; [1.4], pag. 4 "assembling predictions sequentially... NHITS significantly improves accuracy in long-horizon forecasting").
The combination of Shluzas + Olivares makes obvious the full limitation splicing output results of the first model, the second model, the third model and the third model in time and input into the fifth model for integrated training to obtain the health assessment model because a POSITA would implement "stacking, pag. 30" (a meta-model) as a routine configuration to correct the errors of individual base models within the patient monitoring workflow. This is because Shluzas seeks high-accuracy risk prediction for trauma environments and Olivares teaches an integration technique that predictably achieves that goal by leveraging cross-learning across multiple model architectures (Reference Shluzas, See at least, [0105] "scores... are not accurate enough... use basic statistical methods"; Reference Olivares, See at least, [1.1] "global model... share information across them... translates in more accurate forecasts").
A skilled Artisan who read Shluzas would combine Olivares with Shluzas because they both address the technical problem of increasing accuracy in time-series forecasting where single statistical models fail. Shluzas explicitly acknowledges that current "Modified Early Warning Scores, par. 0105" are not accurate enough, and Olivares provides the specific solution by moving beyond simple ensembles to integrated meta-training (Reference Shluzas, See at least, [0103] "combined to provide a more accurate model of actual patient status"; Reference Olivares, See at least, [1.3] "integration of multiple information sources... significantly enhances its accuracy"). This combination resolves the lack of precision in Shluzas' basic aggregation by using the fifth model to learn the complex temporal relationships between the base predictions, resulting in a more robust health assessment score (Reference Olivares, See at least, [1.4] "Assembling predictions sequentially... "). They would have a Reasonable Expectation of Success because Olivares demonstrates that "stacking" diverse models (like ARIMA/Informer into a meta-model) is a "dependable" and "well-proven" method to enhance predictive performance over time.
Shluzas in combination with Narziev + Olivares teaches, Claim 15.
The method according to claim 14, wherein the fifth model is a LightGBM model.
Shluzas teaches an integrated multi-model framework that combines outputs from multiple models (e.g., “aggregating the predictions from multiple weak learners / different models” and “aggregating predictions among voting models… include ‘ADD’ and ‘OR’ aggregation”), which reads on the “integrated” structure into which a later-selected model may be placed. (Shluzas, See at least, [0228] “RNN ensemble method utilizes a voting system”; [0227] “aggregating the predictions from multiple weak learners”).
However, Shluzas does not describe “wherein the fifth model is a LightGBM model”, because Shluzas does not name LightGBM as any model within its disclosed aggregation/ensemble framework.
Olivares teaches the fifth model is a LightGBM model of Claim 15, the missing element in Shluzas that required the selection of this specific gradient boosting decision tree architecture. Olivares explicitly identifies LightGBM as a known, highly efficient model type used in forecasting contexts. (Olivares, See at least, p. 26, par. 3.2.4: “Decision trees... some popular variants include... Gradient Boosting Decision Trees (GBD)... Ke et al., 2017... LightGBM, pag. 26: A highly efficient gradient boosting decision tree”; “These models have consistently performed well in forecasting competitions... and have become well-established baselines”).
A skilled Artisan in the art who read the Shluzas application would combine Olivares with Shluzas, because they are in the same field of endeavor regarding predictive modeling and machine learning ensembles. Shluzas suggests the need for aggregating “different models” to optimize the “assessment of a patient’s condition,” while Olivares provides the specific solution by identifying LightGBM as a “popular variant” of decision trees that excels in such forecasting environments (Shluzas, [0227]; Olivares, p. 26).
The combination of Shluzas + Olivares applications make obvious the full limitation "wherein the fifth model is a LightGBM model" because a POSITA would implement the LightGBM model taught by Olivares as the specific "fifth model" or aggregation layer within the Shluzas ensemble workflow. A POSITA would implement LightGBM as a routine configuration to make the Shluzas integration of "different models" operate with the learned precision of a gradient boosting machine rather than a simple voting system. Because Shluzas seeks the goal of “improved accuracy and sensitivity” through aggregation, and Olivares teaches that LightGBM is a “highly efficient” variant that has “consistently performed well” and is a “well-established baseline” for forecasting tasks, the substitution is a predictable use of a known element to improve the system's predictive capability. (Reference Shluzas + Olivares, See at least, Shluzas: “aggregating the predictions from multiple weak learners... improves accuracy” [0227]; Olivares: “LightGBM: A highly efficient gradient boosting decision tree... have consistently performed well... well-established baselines” [p. 26, par. 3.2.4]).
A skilled artisan would have a Reasonable Expectation of Success because Olivares confirms that LightGBM is a “well-established baseline” that is “effective, easy-to-use” and robust for forecasting tasks. (Reference Olivares, See at least, p. 26, par. 3.2.4: “Tree-based methods are often the selection for users seeking effective, easy-to-use black box learners”).
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