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 .
Continued Examination Under 37 CFR 1.114
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 03 October 2025 has been entered.
Status of the Claims
The amended claim set received 03 October 2025 has been entered into the application.
Claims 1, 8, 12, and 15-17 have been amended.
Claims 2-3, 11, 13-14, and 19 are previously cancelled.
Claims 5-7 and 9-10 cancelled.
Claim 20-25 are new.
Claims 1, 4, 8, 12, 15-18, and 20-25 are pending.
Priority
Applicant claims priority to provisional application 62/715,079 received 06 August 2018 is acknowledged.
Claim Rejections - 35 USC § 101
The instant rejection is maintained for reason for record in the Office Action mailed 14 July 2025 and modified in view of the amendments filed 03 October 2025. It is noted the amendments received 03 October 2025 necessitated new ground(s) of rejection.
The rejection of claims 5-7 and 9-10 under 35 U.S.C. 101 in the Office Action mailed 14 July 2025 is withdrawn in view of the amendments filed 03 October 2025 because the claims were canceled.
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1, 4, 8, 12, 15-18, and 20-25 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
Following the flowchart of MPEP 2106
Step I - Process, Machine, Manufacture or Composition
The claims 1, 4, 8, and 20-22 are drawn to a computing device, so a machine.
Claims 12, 15-18, and 23-25 are drawn to a computer-implemented method, so a process.
2A Prong I - Identification of an Abstract Ideas
Claims 1 drawn is a computing device while claim 12 is drawn to a computer-implemented method. Claims 1 and 12 encompass similar limitations and are examined similarly.
Claims 1 and 12 recite:
obtain gene expression data from one or more gene expression datasets with the gene expression data corresponding to one or more tissue samples
This step can be performed in the human mind to follow instructions to gather gene expression data and is therefore an abstract idea.
segmenting the H&E-stained histopathology image into a plurality of tiles
This step encompasses following instructions to segment the H&E-stained histopathology image into tiles and is therefore an abstract idea. This step encompasses performing mathematical concepts to divide a larger image into smaller more manageable “tiles” using formulas relating to original size, tile size, and overlap which reads on abstract idea. Moreover, it is known in the art that tiling encompasses mathematical formulas such as Total Width = (Num Tiles Across) * (Tile Width), Total Height = (Num Tiles Down) * (Tile Height). To exemplify, in order to tile a 100x100 pixel image to cover a 300x200 area, one would need 3 across (300/100) and 2 down (200/100), making a 3x2 grid.
concatenating and integrating a set of outputs of the gene expression neural network and a set of outputs of the imaging feature neural network in an integrated neural network comprising a third set of neural network layers to (i) learn relationships between the gene expression data and the sets of imaging features. and {ii) produce an integrated neural network output
This step can be performed in the human mind by organizing and combining information (i.e., gene expression and image feature neural network output) to integrate the data into a third neural network layer used for 1) learning relationships between gene expression data and imaging features and 2) to produce an integrated neural network output and is therefore an abstract idea. This step describes the integrated neural network as comprising a third set of neural network layers that 1) learns relationships between gene expression data and image feature and produces am integrated neural network output. This step encompasses using neural network layers to learn relationship and produce output which encompasses mathematical concepts of mathematical relationships between gene expression data and image features and encompasses performing calculations to produce an integrated neural network output which reads on abstract ideas. Here, it is known in the art that neural networks produce output related to mathematical operations (i.e., outputs through interconnected mathematical operations like weighted sums, biases, and activation functions (e.g., sigmoid, ReLU) which reads on abstract ideas. This step encompasses performing mathematical concepts for concatenating and integrating of gene expression and image feature information/data (i.e., output) into a neural network which reads on abstract ideas/mathematical concepts. It is known in the art that concatenating and integrating neural network inputs fundamentally involves extensive mathematics, primarily linear algebra (matrix multiplication, vector addition) and calculus (for training via backpropagation), which allows networks to learn complex patterns from combined data by adjusting weights and biases through weighted sums and nonlinear transformations which reads on abstract ideas. For example, concatenation encompasses vector/matrix mathematics for joining input vectors (features) or outputs from different layers/models into a single, longer vector or larger matrix while integration encompasses mathematically merging information from various streams, often relying on matrix operations and specialized units. Here, the step encompasses concatenation/integrating information (i.e., gene expression data and imaging features), manipulating the data using mathematical functions (i.e., gene expression neural network and image feature neural network), and organizing this information into a new form (i.e., integrated neural network output) which encompasses mathematical concepts. See MPEP 2106.04(a)(2)(A)(iv). Furthermore, and under BRI, the terms “concatenating” and “integrating” are interpreted as combining data.
Claims 4, 16-18, 20, 22-23, and 25 are further drawn to limitations that describe the abstract ideas of claims 1 and 12 and are therefore abstract ideas.
2A Prong II - Consideration of Practical Application.
Claims 1 and 12 do not recite any additional element which integrate the recited judicial exception into a practical application. Here, in the instant case, the claims provide a method of data analysis for inputting information into a neural network to output a set of lymphocyte subgroups. As such, practicing the claims merely results in the inputting information into a neural network to output information which is an extra-solution activity that does not integrate the judicial exception into a practical application.
Additionally, the claims only recite the solution (i.e., identifying lymphocyte subgroups) which attempt to cover any solution for identifying lymphocyte subgroups with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result (i.e., identifying lymphocyte subgroups) which does not integrate the judicial exception into a practical application because this type of recitation is equivalent to the words "apply it". See MPEP 2106.05(f). For example, the claims 1 and 12 recites identifying lymphocyte subtype but claims 1 and 12 and their associated dependent claims do not recite specific genes in order correlating specific lymphocyte subtypes to identify lymphocyte subtypes. To exemplify, claim 18 recites “performing gene selection” but previous and subsequent claims do not encompass or identify any genes associating with any subtypes of lymphocytes. As a further example, claims 1 and 12 and their associated dependent claims do not contain any analysis steps which will be used for identifying lymphocyte subgroups based on gene expression data and image features as claims 1 and 12 are merely drawn to inputting information (i.e., gene expression and image feature data) into neural networks to output identified lymphocyte subtypes.
Furthermore, the neural network is equivalent to merely adding the words “apply it”. Here, the neural networks (NN’s) (i.e., gene expression, image feature, and prediction) are used to generally apply the abstract idea without limiting how the NN’s function. The NN’s are described at a high level such that it amounts to using a computer with a generic NN’s to apply the abstract idea. These limitations do not recite using the NN’s to make predictions but merely recites inputting data into neural network layers to output information. For example, claims 1 and 12 recite the third neural network layers learn relationships and produce output which describes the outcomes of the NN’s without any details about how the outcomes (i.e., learn relationships and produced output, outputting lymphocyte subtypes) are accomplished. See MPEP 2106.05(f).
Thus, such a result only produces information and does not provide for a practical application in the physical-realm of physical things and acts, i.e., the claims do not utilize the data generated by the judicial exception to affect any type of change. See MPEP 2106.04(a)(2)(A)(iv). Therefore, the claims do not utilize the obtained gene expression data and the histopathology image data and the results of the abstract ideas to construct a practical application such as treating a subject, making a tangible object, or improving upon an existing technology.
This judicial exception is not integrated into a practical application because the claims do not meet any of the following criteria:
An additional element reflects an improvement in the functioning of a computer, or an improvement to other technology or technical field;
an additional element that applies or uses a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition;
an additional element implements a judicial exception with, or uses a judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim;
an additional element effects a transformation or reduction of a particular article to a different state or thing; and
an additional element applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than
a drafting effort designed to monopolize the exception.
2B Analysis - Consideration of Additional Elements and Significantly More
The claimed method also recites "additional elements" that are not limitations drawn to an abstract idea.
The recited additional element of using computer process, components, and equipment of claims 1, 12, and 15 does not add significantly more than the recited judicial exception because using computers to analyze abstract ideas is merely tangential to the claimed method and is deemed well-understood, routine, and conventional. See MPEP 2106.05(b), 2106.05(d)(II), and 2106.05(g).
The recited additional element of using extracting data of claims 1, 12, 21, and 24 does not add significantly more than the recited judicial exception because extracting data that is subsequently analyzed by abstract ideas is deemed a well-understood, routine, and conventional extra-solution activity. See MPEP 2106.05(d)(II).
The recited additional element of using data inputting of claims 1, 8 and 12 does not add significantly more than the recited judicial exception because inputting data into a computer for subsequently analysis is merely tangential to the claimed method perform and is deemed well-understood, routine, and conventional extra-solution activity. See MPEP 2106.05(b), 2106.05(d)(II), and 2106.05(g).
The recited additional element of data gathering of claims 1 and 12 does not add significantly more than the recited judicial exception because obtaining histopathology images from tissue is deemed a well-understood and routine insignificant extra-solution activity for providing RNA sequencing data that is subsequently analyzed by the abstract idea. See MPEP 2106.05(d)(II)(i) and (v). To provide evidence of conventionality of obtaining histopathology image and image data, Cooper et al. teaches an example of whole slide imaging and data analysis [page 518 fig 2 and page 519 fig 3] (The Journal of Pathology, 2018-04, Vol.244 (5), p.512-524) (Cited in the Office Action mailed 08 February 2025). To provide further evidence of conventionality, Kothari teaches analyzing whole slide image from ovarian serous carcinoma biopsy [page 1100 figure 2] (Journal of the American Medical Informatics Association: JAMIA, 2013-11) (Cited in the Office Action mailed 08 February 2025).
The recited additional element of data gathering of claims 15 does not add significantly more than the recited judicial exception because using RNA sequencing data source coupled to a communication network is deemed a well-understood and routine insignificant extra-solution activity for providing RNA sequencing data that is subsequently analyzed by the abstract idea. See MPEP 2106.05(d)(II)(i) and (v). To exemplify conventionality of using RNA sequencing data source coupled to a communication network, Hackl et al. (Hackl) computational genomic tools for dissecting tumor-immune cell interactions which reviews utilizing a computer coupled to a next generation sequencing (NGS) machine for processing RNA sequencing data [page 442 box 1] (Nature Reviews Genetics volume 17, pages441–458 (2016) (Cited in the Office Action mailed 08 February 2025)).
In conclusion and when viewed as a whole, these additional claim element(s) do not provide meaningful limitation(s) to transform the abstract idea recited in the instantly presented claims into a patent eligible application of the abstract idea such that the claim(s) amounts to significantly more than the abstract idea itself. Therefore, the claim(s) are rejected under 35 U.S.C. 101 as being directed to non-statutory subject matter.
Response to Arguments
Applicant’s arguments, filed 03 October 2025, have been fully considered but the rejection is maintained.
Law of Nature [page 15 C]
The Applicant states that claims are not directed to a natural law. The Applicant points to Mayo and Cleveland Clinic Foundation for guidance. The Applicant points to amended claims 1 and 12 for clarification. The Applicant further points to Rapid Litig and Illumina Inc v Ariosa. The Applicant states the claims are directed to a new and useful computational technique and therefore, the claims are directed towards a law of nature [remarks, pages 15 -17]. The Applicant states the claims are not directed towards a law of nature. The Applicant points to the Mayo and MPEP 2106.04(b)(I) for guidance. The Applicant points to the amendments to claim 1 for guidance. The Applicant states the multi-stage neural network architecture represents a novel, non-obvious, and concrete technological implementation, not an abstract natural correlation [remarks, page 15]. The Applicant points to both Mayo and Cleveland Clinic for further guidance. The Applicant states “However, in the instant case, existing methods for immune infiltration assessment suffer from significant technical limitations where manual evaluation requires additional tissue samples and pathologist scoring, while existing computational approaches encounter significant ambiguity in reliably identifying correct immune proportions (see paragraph [0005] of the as-filed specification). The claimed multi-stage neural network architecture provides a technological solution by processing gene expression data and histopathological imaging features through separate specialized neural networks before integration, thereby creating a concrete technological framework. Indeed, the Office has not identified any references teaching or suggesting the presently claimed device/method as a whole. Thus, the presently claimed device/method is not well known in the art.” [remarks, page 16-17].
In response, the law of nature rejection is withdrawn because the claims have been amended to recite a method drawn for identifying lymphocytes subset based on general RNA and histopathology image features which does not correlate the natural law with respect to correlating lymphocytes to a patient(s).
Step 2A Prong I [page 17 D]
The Applicant states the claims are not drawn to an abstract idea. The Applicant states Examiners analysis improperly dissects the claims into individual elements and fails to consider the claims are a whole, which recites an improvement that can be practically performed in the human mind [remarks, page 17]. The Applicant states the Examiner’s assertion (FOA)that obtaining gene expression data and segmenting the stained histopathology image into a plurality of tiles can be performed in the human mind ignores the technical reality and scale of the claimed operations. The Applicant points to the specification [0067-0068 and 0077] for guidance. The Applicant states the Examiner has not provided any basis to presume that such tiling and processing sequence can be performed mentally, and that tiling cannot be practically performed in the human mind and represents a concrete improvement. [remarks, page 18].
In response, it is noted that it is known in the art that tiling/segmentation encompasses performing mathematical concepts to divide a larger image into smaller more manageable “tiles” using formulas relating to original size, tile size, and overlap. To exemplify, it is known in the art that tiling encompasses mathematical formulas such as Total Width = (Num Tiles Across) * (Tile Width), Total Height = (Num Tiles Down) * (Tile Height). To exemplify, in order to tile a 100x100 pixel image to cover a 300x200 area, one would need 3 across (300/100) and 2 down (200/100), making a 3x2 grid.
The Applicant states the Examiner’s assertion that a human could extract lymphocyte subtype information by "looking at shapes, dark and light spots, shaded areas or the granularity of an image" improperly suggests that H&E staining alone reveals information that in reality requires specialized IHC staining to visualize. The claimed invention addresses this limitation by computationally integrating H&E imaging features with gene expression data to predict cell compositions that would otherwise require additional tissue and IHC staining.” The Applicant points to the specification [0067-0068 and 0077] for guidance. The Applicant states “The FOA has not provided any basis to presume that such a tiling and processing sequence could be performed mentally, and the FOA's oversimplification sidesteps the technical details recited in the claims. The Applicant states this tiling process cannot practically be performed in the human mind and represents a concrete technical implementation specifically adapted to machine-learning architectures.” [remarks, page 18]. The Applicant states the Examiner oversimplifies the claimed feature extracting process. The Applicant points to specification [0069] for guidance. The Applicant states the human mind cannot practically perform these complex mathematical calculations across hundreds of tiles. [remarks, top of pages 18-19].
In response, and with respect to the invention addressing the computational limitations of integrating H&E imaging features with gene expression data, the argument is not persuasive because combining data, rearranging data, manipulating data to construct/produce integrated/refined data or dataset(s) is still data (i.e., organized information) which reads on organizing data which is an abstract idea. Although the data could be used to address the limitations of integrating image features and gene expression data, here, the claims do not integrate that data with any additional elements such that to construct a practical application or provide an improvement to technology. Furthermore, and under broadest reasonable interpretation (BRI), tiling is interpreted as drawing/overlaying small squares or rectangles over an image of a tissue sample(s). Also, as noted in Step 2A Prong I of the 101 analyses above, tiling encompasses performing mathematical concepts to divide a larger image into smaller more manageable “tiles” using formulas relating to original size, tile size, and overlap which reads on abstract idea. Furthermore, the claims utilize tangential computer elements for carrying out abstract ideas (i.e., segmenting an image into tiles) in a computer environment. See MPEP 2106.04(III)(C)(1-3). Therefore, using tiling limitations does not provide a concrete technical implementation specifically adapted to machine-learning architectures because segmenting an image into tiles reads on abstract ideas. Additionally, extracting information is an additional element that is evaluated under Step 2A Prong II and 2B of the 101 analyses. See MPEP 2106.05(d)(II).
Moreover, and with respect to the human mind not being able to practically perform complex mathematical calculations across hundreds of tiles, the argument is not persuasive because it is acknowledged that such computations (i.e., tiling calculations) performed mentally, or with paper and pencil, would take considerable time and effort, but that is, of course, the singular purpose of computers and computer networks, to perform large numbers of calculations, via algorithms, rapidly, and without error (assuming no error in user input). Although a general-purpose computer can perform calculations at a rate and accuracy that can far outstrip the mental performance of a skilled artisan, the nature of the activity is essentially the same, and constitutes an abstract idea. See SiRF Tech: "In order for the addition of a machine to impose a meaningful limit on the scope of a claim, it must play a significant part in permitting the claimed method to be performed, rather than function solely as an obvious mechanism for permitting a solution to be achieved more quickly, i.e., through the utilization of a computer for performing calculations" and Bancorp: "the fact that the required calculations could be performed more efficiently via a computer does not materially alter the patent eligibility of the claimed subject matter. … Using a computer to accelerate an ineligible mental process does not make that process patent-eligible".
The Applicant states the limits of genomic data processing a human cannot mentally deconvolve bulk RNA sequencing results to extract the cell-type-level features that the claims address. The Applicant points to the specification [0005] for guidance. The Applicant states "gene expression neural network comprising a first set of neural network layers" performs computational deconvolution of bulk gene expression data to learn representations relevant to lymphocyte subtype identification.” This step is inherently computational and cannot be performed by mental processes. [remarks, page 19].
In response, it is noted that the method of claim 1 and 12 are computer implemented methods. Here, the claimed method implements a computer system for analyzing nucleic acid and image data. As such, using a computer to process and evaluate RNA data is equivalent to performing mental processes on a generic computer, performing mental processes in a computing environment, and using a computer a tool to perform a mental process. See MPEP 2106.04(a)(2)(III)(C). With respect to the computational deconvolution of bulk gene expression data to learn representations relevant to lymphocyte subtype identification, the argument is not persuasive because it is noted the claims do not recite methods or steps for deconvolution of bulk gene expression data. Additionally, in light of the specification, deconvolution of RNA sequencing data (i.e., estimating the proportions of different cell types within a mixed (bulk) tissue sample) encompasses mathematical concepts such as regression-based techniques [page 20 para 0085] which reads on abstract ideas/mathematical concepts. As such, and as noted above, the claims utilize computer elements for processing the claimed method. Therefore, performing deconvolution methods (i.e., calculating/estimating the proportions of different cell types within a mixed tissue sample) reads on abstract ideas performed in a computer environment. See MPEP 2106.04(a)(2)(III)(C).
The Applicant states the Examiner never explains how a human could combine the two independent data sets in the manner claimed. The Applicant points to amended claim 1 concatenating step for guidance. The Applicant further points to the specification [0078] for guidance. The Applicant states the process cannot be performed in the human mind. The Applicant states “A proper mental process rejection requires identifying what the human's "inputs" and "outputs" would look like at each stage. Here, the FOA never describes what a human would be looking at after visually observing the H&E-stained image, what they would do with the bulk RNA sequencing data, or how they could combine the two in the equivalent of an integration layer comprising neural network layers that learns relationships between the modalities.” [remarks, pages 19-20].
In response, and with respect to proper mental process rejection, Examiners are to follow the MPEP 2106.07(a) for formulating rejections with respect to identifying mental processes in Step 2A Prong I of the 101 analyses. With respect to concatenating and integrating data, it is known in the art that “concatenating and integrating neural network inputs fundamentally involves extensive mathematics, primarily linear algebra (matrix multiplication, vector addition) and calculus (for training via backpropagation), which allows networks to learn complex patterns from combined data by adjusting weights and biases through weighted sums and nonlinear transformations” which reads on abstract ideas. For example, concatenation encompasses vector/matrix mathematics for joining input vectors (features) or outputs from different layers/models into a single, longer vector or larger matrix while integration encompasses mathematically merging information from various streams, often relying on matrix operations and specialized units. Furthermore, the step encompasses taking information (i.e., gene expression data and imaging features), manipulating the data using mathematical functions (i.e., gene expression neural network and image feature neural network), and organizing this information into a new form (i.e., integrated neural network output) which encompasses mathematical concepts. See MPEP 2106.04(a)(2)(A)(iv).
The Applicant states “the FOA's interpretation of "integrate" as merely "combining data" oversimplifies the claimed integration process. The specification establishes that "neural networks can function as flexible and efficient learning models when integrating heterogeneous features, such as gene expression and imaging features" and that "the machine learning framework uses a neural network-based architecture to integrate RNA seq and imaging data" (see paragraph [0072]). This is not simple data combination but rather a sophisticated machine learning process for finding complex relationships between different biological data modalities.” [remarks, page 20].
In response, it is known in the art that neural networks produce output related to mathematical operations (i.e., outputs through interconnected mathematical operations like weighted sums, biases, and activation functions (e.g., sigmoid, ReLU) which reads on abstract ideas. Additionally, it is known that concatenating and integrating neural network inputs fundamentally involves extensive mathematics, primarily linear algebra (matrix multiplication, vector addition) and calculus (for training via backpropagation), which allows networks to learn complex patterns from combined data by adjusting weights and biases through weighted sums and nonlinear transformations which also reads on abstract ideas. For example, concatenation encompasses vector/matrix mathematics for joining input vectors (features) or outputs from different layers/models into a single, longer vector or larger matrix while integration encompasses mathematically merging information from various streams, often relying on matrix operations and specialized units. Here, the concatenating and integrating steps read on abstract ideas as noted in Step 2A Prong I of the 101 analyses above.
The Applicant states the claimed invention represents an important advancement invention in elaborating the tumor microenvironment and in predicting the immunological composition of individual patients.”. The Applicant states improvements to computational pathology technology cannot be performed and achieved through mental process alone. [remarks, page 20]. The Applicant states “Accordingly, when properly considered as a whole, the claims are not directed to mental processes or mathematical concepts that can be performed in the human mind, but rather recite a specific technological implementation that improves computational pathology technology [remarks, page 21].
In response and as noted above, segmenting the H&E-stained histopathology image into a plurality of tile encompasses performing mathematical concepts to divide a larger image into smaller more manageable “tiles” using formulas relating to original size, tile size, and overlap which reads on abstract idea. On the other hand, concatenating and integrating a set of outputs can also be performed in the human mind by organizing and combining information (i.e., gene expression and image feature neural network output) to integrate the data into a third neural network layer. Furthermore, this step describes the integrated neural network as comprising a third set of neural network layers that 1) learns relationships between gene expression data and image feature and produces am integrated neural network output. Here, it is known in the art that neural networks produce output related to mathematical operations (i.e., outputs through interconnected mathematical operations like weighted sums, biases, and activation functions (e.g., sigmoid, ReLU) which reads on abstract ideas. Moreover, it is known in the art that “concatenating and integrating neural network inputs fundamentally involves extensive mathematics, primarily linear algebra (matrix multiplication, vector addition) and calculus (for training via backpropagation), which allows networks to learn complex patterns from combined data by adjusting weights and biases through weighted sums and nonlinear transformations” which also reads on abstract ideas. For example, concatenation encompasses vector/matrix mathematics for joining input vectors (features) or outputs from different layers/models into a single, longer vector or larger matrix while integration encompasses mathematically merging information from various streams, often relying on matrix operations and specialized units. Here, the step encompasses concatenation/integrating information (i.e., gene expression data and imaging features), manipulating the data using mathematical functions (i.e., gene expression neural network and image feature neural network), and organizing this information into a new form (i.e., integrated neural network output) which encompasses mathematical concepts. See MPEP 2106.04(a)(2)(A)(iv). It is further noted that determining whether a claimed method recites improvements is evaluated under Step 2A Prong II of the 101 analyses.
Step 2A Prong II [page 21 E]
The Applicant states the claim are not directed to a judicial exception under Prong II of Step 2A because the claims integrate the judicial exception into a practical application. The Applicant points to the specification [0005] for guidance. The Applicant points to the amendments to claim 1 for guidance. The Applicant states “This four-stage neural network architecture provides a technological solution that integrates heterogeneous biological data types in a manner that was not previously possible.” [remarks, pages 21-22]. The Applicant points to the specification [0040 and 0088] for further guidance. The Applicant states “the claims thus integrate any judicial exception into a practical application by providing a concrete technological improvement to computational pathology that replaces manual, resource-intensive evaluation methods with an automated, multimodal analysis framework that achieves superior accuracy. This constitutes a practical application in the technological field of computational pathology, not merely the production of abstract information.” [remarks, page 22].
In response, and as noted in Step 2A Prong II of the 101 analyses above, the claims do not contain any additional elements that integrate judicial exception into a practical application. With respect to the four-stage neural network architecture providing a technological solution, it is noted that a neural network has at least two layers (input and output), but typically includes one or more hidden layers, with "deep" networks having many hidden layers to learn complex patterns, ranging from a few to hundreds, depending on the task's complexity. The number of layers isn't fixed and is determined by the problem, with more layers enabling deeper, more abstract feature learning. Thus, the determined problem of the instant claims requires using four layers (i.e., gene expression, image features, integrated output, prediction output) which does not construct an architecture that provides a technological solution (i.e., improvement to technology) because it is acknowledged that layers of a neural network are not fixed and is dependent on the required task. Therefore, the implementation of four layers does not provide a practical application of an improved neural network architecture for identifying lymphocyte subgroups.
Moreover, with respect to the neural network (NN), it is recited at a high level and reads on mathematical processes such as vector and matrix mathematics as explained in Step 2A Prong I of the 101 analyses above. Furthermore, the NN reads on any simple neural network which encompasses simple matrix/matrices. Additionally, in view of the specification, the specification does not provide support/evidence that the claimed invention provides an improvement to a technological field. Here, the specification contains conclusionary statements without accompanying adequate support (i.e., quantitative and qualitative) the claimed method is concrete technological improvement to computational pathology that replaces manual, resource-intensive evaluation methods with an automated, multimodal analysis framework that achieves superior accuracy. Furthermore, with respect to the neural network, the neural network is recited a high level of generality and therefore reads on abstract ideas/mathematical concepts. Therefore, the additional elements (i.e., computer elements, data gathering and data inputting) do not integrate the recited judicial exception into a practical application.
The argument is not further persuasive because as noted in Step 2A Prong II of the 101 analyses above, the neural network is at best equivalent to merely adding the words “apply it”. Similarly, it is noted that this rationale is also applied if the neural network is considered as an additional element which would also at best be equivalent to merely adding the words “apply it”. Here, the neural networks (NN’s) (i.e., gene expression, image feature, and prediction) are used to apply the idea without limiting how the NN’s function. Here, the NN’s are described at a high level such that it amounts to using a computer with a generic ANN to apply the abstract idea. It is noted these limitations do not recite using the NN’s but merely recites inputting data into neural network layers. See 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence USPTO Example 47 Claim 2. Also, claims 1 and 12 recite the third neural network layers to learn relationships and produce output that is subsequently inputted into a fourth neural network layer which describes the outcomes of the NN’s without any details about how the outcomes (i.e., learn relationships and produced output, outputting lymphocyte subtypes) are accomplished. See MPEP 2106.05(f).
The Applicant states the claims provide a concrete technological improvement to lymphocyte subtype identification through its specific multi-stage neural network architecture which provides a custom neural network architecture by providing a concrete technological improvement to computational pathology using a neural network architecture. The Applicant states the use of tile-based image segmentation, extraction of imaging features, and the integration of separate neural networks for specific data types is not routine or generic. The Applicant states the processes is performed on a computing device that cannot be performed in the human mind or mental steps [remarks, pages 22, third paragraph].
In response and as noted in Step 2A Prong II of the 101 analyses above, the neural network is recited as high level of generality and reads on abstract ideas (i.e., vector and matrix mathematics) and amounts to implementing abstract ideas on a computer, or merely uses a computer as a tool to perform the abstract ideas. See MPEP 2106.05(f). With respect to the integration of separate neural networks for specific data types is not routine or generic, routineness and conventionality are evaluated under Step 2B of the 101 analyses. As such, the steps are not applied to any additional elements so as to result in a practical application or an improvement to technology. Novel or improved abstract idea steps alone are not deemed to be an improvement to technology. Furthermore, abstract ideas are evaluated under Step 2A Prong I of the 101 analyses, not in Step 2A Prong II. Additionally, tiling reads on abstract ideas (i.e., mathematic concepts) as noted in Step 2A Prong I of the 101 analyses above.
The Applicant states the claimed multi-stage architecture represents a departure from conventional approaches. The Applicant points to the amendments to claim 1 for guidance. The Applicant states the amendments to claim 1 “enable for specialized processing of each input type using different feature extraction methodologies tailored to the unique characteristics of gene expression data versus histopathological imaging data. This specialized processing is fundamentally different from generic data analysis because it implements domain-specific feature extraction techniques optimized for each biological data type.” [remarks, page 23]. The Applicant states “the concatenating integrating step preserve the unique characteristics of each data modality while enabling cross-modal learning. This concatenation-based integration approach enables the claims to weight the unique findings from each specialized neural network pathway, creating a custom weighting of feature importance, thereby enabling the system to automatically discover optimal combinations of gene expression and imaging features for immune infiltration prediction.” [remarks, page 23]. The Applicant states the prediction neural network represents a technological advancement by “inputting” data into a fourth neural network. The Applicant points to the specification [0060] for guidance. The Applicant states the claims provides a practical application that improves the technological field of computational pathology. [remarks, page 23].
In response, it is noted that conventionality is evaluated under Step 2B of the 101 analyses. With respect to concatenating and integrating data, it is known that concatenating and integrating neural network inputs fundamentally involves extensive mathematics, primarily linear algebra (matrix multiplication, vector addition) and calculus (for training via backpropagation), which allows networks to learn complex patterns from combined data by adjusting weights and biases through weighted sums and nonlinear transformations which reads on abstract ideas. For example, concatenation encompasses vector/matrix mathematics for joining input vectors (features) or outputs from different layers/models into a single, longer vector or larger matrix while integration encompasses mathematically merging information from various streams, often relying on matrix operations and specialized units. With respect to the specification [0060] providing evidence that integration of imaging features greatly improves prediction of the immune infiltrate in comparison to conventional techniques, as the features exhibit improved efficacy across various cell types, it is noted that the paragraph provides a mere conclusionary statement related to immune infiltration score being a percentage of predicted immune cells which does not provide evidence demonstrating that integration of imaging features greatly improves prediction of the immune infiltrate in comparison to conventional techniques, as the features exhibit improved efficacy across various cell types. With respect to being different from generic data analysis because it implements domain-specific feature extraction techniques optimized for each biological data type and the concatenating integrating step preserve the unique characteristics of each data modality while enabling cross-modal learning, it is noted that these features are not recited in the claims. Furthermore, these features describing a specialized neural network (NN) structure are not recited in the claims. The claims are directed to a multi-step method of merely using generic neural networks at each of the steps to further analyze the output of a previous neural network, or to further analyze the combined outputted data from two previous neural networks. Here the neural network is recited at a high level of generality and reads on math (e.g., if considered as a simple NN) or a generic computer (e.g., if considered as an additional element). Thus, in light of the specification the claimed methods do not recite a practical application and/or an improvement to technology. Therefore, the claims are not patent eligible under Step 2A Prong II of the 101 analyses.
2B [page 24 F]
The Applicant states the claims amount to significantly more than the recited judicial exception. [remarks, page 24]. The Applicant points to claim 1 inputting gene expression data, inputting image feature data, concatenating and integrating data and inputting data into a fourth set of neural network layers. The Applicant states the specification architecture does beyond generic computer implementation by providing concrete structural limitations that define how heterogeneous biological data types are processed through separate specialized pathways before integration [remarks, page 24].
In response, the argument is not persuasive because the claimed steps, individually and as a whole, do not provide significantly because inputting data is a well-known and conventional additional for providing data that is subsequently analyzed or acted upon by the abstract idea. Additionally, concatenating and integrating data is considered an abstract as described in Step 2A Prong I of the 101 analyses above.
The Applicant states the specification supports nonconventional nature of image processing. The Applicant points to the specification [0067-0069] for guidance. The Applicant states “This level of technical specificity in processing H&E-stained histopathological images through tile-based segmentation and multi-modal feature extraction represents a non-conventional application that cannot be performed by mental processes or generic computer implementation.” [remarks, pages 24-25].
In response, the argument is not persuasive because abstract ideas are evaluated under Step 2A Prong I of the 101 analyses. Furthermore, segmenting an image into tiles encompasses performing math and reads on abstract ideas as noted in the Step 2A Prong I of the 101 analyses above. Furthermore, extracting data is deemed a well-known and conventional insignificant extra-solution activity. See MPEP 2106.05(d)(II)(v).
The Applicant states specification provides a concrete neural network architecture. The Applicant points to the specification [0076-0078] for guidance. The Applicant states “the claims address the technical challenge of accurately identifying lymphocyte subtypes in tissue samples by integrating heterogeneous data sources (namely, gene expression data and histopathology images) using a custom neural network architecture. The use of tile-based image segmentation, extraction of certain imaging features, and the integration of separate neural networks for each data type is not routine or generic.” [remarks, page 26]. The Applicant points to the specification [0040 and 0088] for guidance. The Applicant states the claims contains additional elements, both individually and in combination, that transform any alleged abstract idea into a patent-eligible application, and thus the claims satisfy Step 2B of the subject matter eligibility test.” [remarks, page 26].
In response, the claims encompass routine and conventional additional elements which does not provide significantly more because the claims do not encompass additional elements outside of tangential computer elements, nucleic acid sequencing, and data inputting and outputting. Here, the claimed steps recite obtaining data (i.e., gene expression data and image feature data), extracting data, inputting data into a neural network, and producing integrated neural network output which encompasses routine and conventional additional elements. Furthermore, with respect to the neural networks, the limitations are evaluated under Step 2A Prong I of the 101 analyses above as encompassing abstract ideas. As such, the claims are drawn to gathering and analyzing data (i.e., gene expression and image feature data) using conventional techniques (i.e., computer elements and nucleic acid sequencing) for inputting the results into a prediction neural network which is insufficient to be patient eligible under Step 2B of the 101 analyses because the additional elements do not provide significantly more than the recited judicial exception.
Conclusion
Claims 1, 4, 8, 12, 15-18, and 20-25 are rejected.
No claims are allowed.
Finality
This Office action is a Non-Final action. A shortened statutory period for reply to this action is set to expire THREE MONTHS from the mailing date of this action.
Inquiries
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOSEPH C PULLIAM whose telephone number is (571)272-8696. The examiner can normally be reached 0730-1700 M-F.
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, Karlheinz Skowronek can be reached at (571) 272-9047. 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.
/J.C.P./ Examiner, Art Unit 1687
/Anna Skibinsky/
Primary Examiner, AU 1635