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 .
Response to Amendment
The Amendment filed on July 20th, 2026, has been entered. Claims 36–40 are currently pending. Claims 36–39 have been amended and claim 40 remains previously presented.
Response to Arguments
Applicants’ arguments filed 07/20/2026 have been fully considered. Applicant’s arguments are persuasive in PART with respect to the previous claim objection, the previous rejections of claims 36–40 under 35 U.S.C. § 103, and not persuasive with respect to the previous rejection of claims 36–40 under 35 U.S.C. § 101 as explained below.
Applicant’s arguments, see page 11 of the Remarks, state that claims 36 and 38 have been amended to replace the previously unclear “posterior subcapsular of a lens” terminology with “posterior subcapsular cataract.” The Examiner agrees that the amendment clarifies the intended terminology. Therefore, the previous objection to claim 36 has been withdrawn.
Applicant's arguments, see pages 11–13 of the Remarks, state that amended claim 36 is no longer directed to a judicial exception because the newly added limitation, "wherein the deep learning prediction model uses, as an objective function, a function obtained by combining a Class Balanced (CB) loss function and a Generalized Cross Entropy (GCE) loss function", recites a specific technical configuration of the model itself, improving the model's accuracy despite data imbalance and robustness despite incorrect labeling, and thereby integrates the claim into a practical application and supplies significantly more under Step 2B. The Examiner has fully considered this argument but does not find it persuasive. Because both individual loss-function components, and their combination into a single objective function addressing both class-imbalance and label-noise problems, were each independently known in the prior art before the effective filing date, the newly added limitation of amended claim 36 reflects the routine, conventional application of known techniques to a known problem, rather than an unconventional or specific improvement to the operation, architecture, or training methodology of the deep learning model itself, as contemplated by Ex parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025), and does not distinguish this case from Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1178 (Fed. Cir. 2025). Accordingly, the rejection of claims 36–40 under 35 U.S.C. § 101 is maintained.
Applicant’s arguments, see pages 14–16 of the Remarks, state that neither Wu nor Gagnon teaches or suggests the newly added limitation requiring the deep learning prediction model to use an objective function obtained by combining a CB loss function and a GCE loss function. Applicant further argues that Jiang and Chylack do not cure this deficiency for dependent claims 37–40. The Examiner agrees that the references relied upon in the previous rejections do not expressly teach the newly added combined CB/GCE objective-function limitation. Because claims 37–40 directly or indirectly depend from amended claim 36 and therefore incorporate the newly added limitation, the previous rejections of claims 36–40 under 35 U.S.C. § 103 have been withdrawn in view of the updated rejection as fully disclosed below.
Further searching directed to the newly added objective-function limitation identified two foundational pre-priority publications: Zhang et al, “Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels” (2018), which teaches GCE loss for improving deep-network robustness in noisy-label environments, and Cui et al., “Class-Balanced Loss Based on Effective Number of Samples” (2019), which teaches class-balanced loss based on effective sample counts to address class-imbalanced training data. These references separately teach the two loss-function components and the respective technical problems identified in Applicant’s specification: incorrect or noisy labels and class imbalance. During the continued search, Li et al., CN 111797703 A, resurfaced as the reference addressing the claimed combination, expressly discloses a single robust loss function, denoted RSRLF, obtained by multiplying a Generalized Cross-Entropy-type fault-tolerant loss function by an adaptive class-balance weight, together with an express rationale for combining the two: that the fault-tolerant (GCE) component suppresses learning from noisy/low-confidence samples while the adaptive class-balance component promotes learning of underrepresented classes, and that jointly optimizing the two avoids the shortcomings of using either technique alone. Li expressly teaches an RSRLF robust objective function defined as (
L
R
S
R
L
F
=
L
G
C
E
×
w
k
), wherein (where
L
G
C
E
) is a Generalized Cross Entropy loss function and (
w
k
) is an adaptive class-balance weight. Li further explains that the GCE component improves robustness to noisy labels, while the adaptive class-balance component addresses differences in class-sample size, and reports improved classification performance from their joint use. Accordingly, Li is applied in this Office Action to address the combined CB/GCE objective-function limitation added to claim 36.
The need to apply Li and replace the previous grounds of rejection was directly necessitated by Applicant’s substantive amendment adding the combined CB/GCE objective-function limitation to independent claim 36, which is inherited by dependent claims 37–40. Applicants’ amendment therefore necessitated the new grounds of rejection presented in this Office Action. Accordingly, the new grounds of rejection were necessitated by Applicant’s amendments, and this action is properly made final in accordance with MPEP § 706.07(a).
Based on these facts, this action is made FINAL.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 36–40 are rejected under 35 U.S.C. § 101 because the claimed invention is directed to a judicial exception without significantly more.
This rejection has been made in accordance with the current USPTO subject matter eligibility framework, including MPEP §§ 2103–2106.07, the 2019 Revised Patent Subject Matter Eligibility Guidance, the October 2019 Patent Eligibility Guidance Update, the 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence, the July 2024 AI Subject Matter Eligibility Examples, the August 4, 2025 USPTO memorandum titled "Reminders on evaluating subject matter eligibility of claims under 35 U.S.C. § 101," and the USPTO's guidance concerning Ex parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025). The claims have been evaluated under the broadest reasonable interpretation, and the claims have been considered as a whole, including the newly added limitation of amended claim 36 discussed below.
Step 1: Statutory category
Independent claim 36 is directed to a method for diagnosis of cataract using deep learning and therefore falls within the statutory category of a process. Claims 37–40 depend therefrom. Accordingly, the analysis proceeds to Step 2A.
Step 2A, Prong One (Judicial exception)
Independent claim 36 recites: receiving a slit lamp microscopic examination result image, a retroillumination examination result image, and a medical examination result of a subject; inputting the images into a deep learning prediction model to estimate degrees of cataract progression in a nucleus, a cortex, and a posterior subcapsular cataract, and determining whether the subject has cataract; evaluating cataract severity based on the degrees of cataract progression; and determining and providing a necessary treatment stage using the cataract severity or the medical examination result.
These limitations recite observations, evaluations, judgments, and recommendations that describe a diagnostic decision-making process of the type practically performed in the human mind by a doctor, with the aid of pen and paper, and therefore fall within the mental-process grouping of abstract ideas. Receiving ophthalmic images and a medical result, evaluating the degree of opacity present in different regions of the lens, concluding whether cataract is present, judging its severity, and recommending a course of action (follow-up, hospital visit, or surgery) are the type of clinical observation, evaluation, and judgment steps a treating physician performs when reviewing slit-lamp and retroillumination photographs and a patient's visual acuity result. The recited "deep learning prediction model" does not remove the claim from the abstract-idea grouping because it is invoked functionally to carry out the diagnostic evaluation, rather than being claimed as a specific improvement to the model's own operation, architecture, or training (see MPEP § 2106.04(a)(2)).
Claim 36 further recites, as amended, "wherein the deep learning prediction model uses, as an objective function, a function obtained by combining a Class Balanced (CB) loss function and a Generalized Cross Entropy (GCE) loss function." This limitation independently recites a mathematical concept; specifically, a mathematical formula combining two known error-computation functions to numerically evaluate and adjust model parameters during training. Reciting that a loss function is "obtained by combining" two named mathematical loss functions is, in substance, the recitation of a mathematical relationship or formula, analogous to claims reciting loss-function optimization or weighted mathematical combination of terms, which fall within the mathematical-concepts grouping of abstract ideas.
Claims 37–40 further limit the method to lens-region extraction using Faster R-CNN, particular NO/ NC/ CO/ PSC grading logic, visual-acuity-based severity categorization, and specific recommendation outputs (follow-up, hospital visit, surgery). These limitations further define the same mental evaluative process: identifying, grading, and acting on lens opacity findings and do not remove the claim from the abstract-idea groupings identified above.
Accordingly, claims 36–40 recite an abstract idea under Step 2A, Prong One.
Step 2A, Prong Two (Practical Application)
The additional elements, considered individually and in combination, do not integrate the abstract idea into a practical application.
Receiving the slit-lamp image, retroillumination image, and medical examination result amounts to data gathering for use in the abstract diagnostic analysis. Determining and providing a treatment stage is merely outputting the result of that analysis, not administering treatment or controlling a medical device. Limiting the abstract idea to the field of cataract diagnosis and to particular ophthalmic image modalities is a limitation to a particular technological environment rather than a practical application of the exception. Claim 37's recitation of Faster R-CNN lens-region extraction adds only generic preprocessing of input data before the same diagnostic analysis. Claims 38–40's grading, severity, and recommendation logic remain within the same abstract evaluative process.
With respect to amended claim 36's newly added objective-function limitation, the Examiner has evaluated whether this limitation reflects a specific technological improvement to the deep learning model itself, as opposed to the routine application of an existing mathematical technique in a particular field of use. In evaluating this question, the Examiner identified two foundational, pre-priority-date references, each addressing one of the two component technical problems Applicant's own specification attributes to the recited loss functions. Zhang et al, "Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels" (NeurIPS 2018), teaches a Generalized Cross Entropy loss function developed specifically to provide robustness to noisy or incorrectly-labeled training data. Cui et al., "Class-Balanced Loss Based on Effective Number of Samples" (CVPR 2019), teaches a class-balanced loss re-weighting scheme, computed from the effective number of training samples per class, developed specifically to correct for class-imbalance in training data; Cui expressly characterizes this re-weighting term as model-agnostic and loss-agnostic, i.e., applicable atop any existing base loss function. The Examiner's search then determined whether a pre–July 22, 2021 reference expressly combines the two, and identified Li et al., CN 111797703 A (published October 20, 2020), which discloses a single robust loss function obtained by multiplying a Generalized-Cross-Entropy-type fault-tolerant loss by an adaptive class-balance weight, together with an express rationale that the fault-tolerant component suppresses overfitting to noisy samples while the class-balance component promotes learning of underrepresented classes.
Because both individual loss-function components, and their combination into a single objective function addressing both class-imbalance and label-noise problems, were each independently known in the prior art before the effective filing date, the newly added limitation of amended claim 36 reflects the routine, conventional application of known techniques to a known problem, rather than an unconventional or specific improvement to the operation, architecture, or training methodology of the deep learning model itself, as contemplated by Ex parte Desjardins, Appeal No. 2024-000567 (PTAB Sept. 26, 2025). This case is distinguishable from Desjardins, where the claims reflected disclosed improvements in how the machine-learning model itself operated, including reduced storage, reduced system complexity/streamlining, and preservation of performance on prior tasks during subsequent learning. Here, by contrast, the claims apply two independently known, off-the-shelf loss functions, already demonstrated in combination in the prior art for an analogous class-imbalance/label-noise problem in a different image-classification/segmentation context, to the field of cataract diagnosis. This is consistent with Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1178 (Fed. Cir. 2025), where the Federal Circuit explained that claims directed to using a generic machine-learning technique in a particular environment are abstract when they do not claim the machine-learning improvement itself; applying a known combined loss-function technique to a new field of use (cataract diagnosis, in place of Li's remote-sensing classification) is a field-of-use limitation, not an integration of the abstract idea into a practical application.
Accordingly, claims 36–40 do not integrate the judicial exception into a practical application under Step 2A, Prong Two.
Step 2B: (Inventive Concept)
The additional elements, considered both individually and as an ordered combination, do not amount to significantly more than the abstract idea.
The claims use generic computer/ AI elements performing their ordinary functions of receiving data, preprocessing data, analyzing data via a deep learning model, and reporting a result. As discussed above, both the Class-Balanced loss component and the Generalized Cross Entropy loss component of the recited objective function were independently known conventional techniques prior to the effective filing date, and their combination into a single objective function addressing the same two technical problems (class imbalance and label noise) was also known prior to the effective filing date. Considered individually and in ordered combination, the additional elements are therefore the well-understood, routine, and conventional application of known mathematical loss-function techniques, and do not amount to significantly more than the abstract idea itself.
Claim 37's Faster R-CNN extraction is generic preprocessing of image inputs. Claim 38's NO/NC/CO/PSC grading and cataract/no-cataract decision logic merely further define the abstract medical evaluation. Claim 39's categorization into non-cataract, mild, moderate, and severe using visual acuity relative to a threshold merely further defines the evaluative rules used in the diagnosis/severity assessment. Claim 40's output of recommendations such as follow-up examination, hospital visit, or surgery merely reports the result of the abstract evaluation and does not itself perform treatment. These limitations do not add a specific technical improvement or other element sufficient to amount to significantly more than the abstract idea.
Accordingly, claims 36–40 are not directed to patent-eligible subject matter and are therefore rejected under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim 36 is rejected under 35 U.S.C. §103 as being unpatentable over Wu (Wu et al, Universal Artificial Intelligence Platform for collaborative management of Cataracts, 2019) in view of Gagnon (Gagnon et al, US 2021/0127967 A1, 2021) and Li (Li et al., CN 111797703 A).
Regarding claim 36, Wu teaches a method for diagnosis of cataract using deep learning ( Wu, in [“Introduction”], discloses a universal AI plat form for the collaborative diagnosis and work-flow management of cataracts. ), the method comprising:
a step of input in which an input unit receives a slit lamp microscopic examination result image,
( Wu, in [“Method - Cataract diagnosis and management model for the AI agent”,], teaches that, in step 1, the derived slit lamp photographs are classified into four separate capture modes; also in [Fig. 5], images are obtained by slit lamp microscopes; the overall cataract AI workflow teaches receiving a slit-lamp microscopic examination result image as input to the model. )
a step of diagnosing cataract in which a cataract diagnosing unit inputs the slit lamp microscopic examination result image, the
( Wu, in [“Cataract diagnosis and management model for the AI agent”], teaches that, in step 2, each photograph was diagnosed as a normal lens, cataract or a postoperative eye, and, in step 3, aetiological classification and cataract severity were used to further subclassify each diagnosed photograph with respect to a management strategy; in [“Deep learning convolutional neural network for training and classification”] further teaches that ResNet was used and that mode recognition (type of cataract, such as: a nucleus, a cortex and a posterior subcapsular of a lens), cataract diagnosis, and severity evaluation were trained as separate deep learning tasks. )
wherein the deep learning prediction model is trained using stochastic gradient descent, with data augmentation performed before training to balance data from different categories, and uses a softmax output layer to select the category corresponding to the maximum output value;
( Wu, in [“Deep learning convolutional neural network for training and classification”], [p. 1554]: teaches that stochastic gradient descent was used to train the ResNet network; data augmentation was performed before training through random rotation, translation, cropping, and flipping to balance data from different categories; and a softmax layer was directly used to select the category corresponding to the maximum output value. )
a step of evaluating cataract severity in which a severity evaluating unit evaluates the cataract severity of the subject based on the degrees of cataract progression in the nucleus, the cortex and the posterior subcapsular cataract; and
( Wu, in [“Cataract diagnosis and management model for the AI agent”] teaches that, in step 3, following aetiological classification and cataract severity were used to further subclassify each diagnosed photograph with respect to a management strategy of referral or follow-up, that the severity of adult cataracts was evaluated primarily by LOCS II nuclear grades (I-IV), with III or IV defined as severe / referral and I-II defined as mild, and with secondary evaluation for significant posterior capsular opacification (PCO) or anterior capsular opacification (ACO);. and in [Figure 4, "Results"], further teaches adult cataract severity evaluation and detection of referable subcapsular opacification. )
a step of providing treatment plan in which a treatment stage determining unit determines and provides a necessary treatment stage of cataract for the subject using the cataract severity or the medical examination result.
( Wu, in [Figure 4, "Results"], teaches a novel tertiary healthcare referral system in which VA and a brief case history are collected, suspicious cases are referred to community healthcare facilities where anterior segment images are obtained, the cataract AI agent provides a comprehensive evaluation considering diagnosis and referable conditions, and if the AI agent decides that the cataract is a referral, a fast-track notification system is triggered and patients are informed to undergo a comprehensive examination. )
As noted above in square brackets, Wu does not expressly disclose where Gagnon teaches:
a step of input in which an input unit receives a slit lamp microscopic examination result image, a retroillumination examination result image and a medical examination result of a subject;
( Gagnon, in [0015], [0059-0060], [0078-0087] and [0111], teaches an automated slit-lamp method / system in which the controller sequences different illumination patterns selected from a group including retroillumination and the camera generates corresponding image sets; imaging records are stored on a memory for later diagnostics; and further teaches communication between the automated slit lamp system and one or more remote databases comprising an electronic medical record of the patient. )
a step of diagnosing cataract in which a cataract diagnosing unit inputs the slit lamp microscopic examination result image, the retroillumination examination result image into a deep learning prediction model
( Gagnon, in [0087], teaches the dual-image input side and using machine-learning techniques to output diagnostic data based on the acquired images. )
It would have been obvious to modify Wu to use the additional retroillumination image input and medical result input taught by Gagnon, because Wu already applies deep learning to cataract diagnosis, severity assessment, and referral management, and Gagnon teaches acquiring multiple slit-lamp image sets including retroillumination and communicating with patient electronic medical records. Combining Gagnon’s known input-acquisition features with Wu’s known AI cataract workflow would have been a predictable use of prior-art elements according to their established functions to obtain cataract diagnosis and management from a more complete set of ophthalmic image and medical-result inputs.
Wu [as modified by Gagnon] does not expressly disclose an objective function obtained by combining a Class-Balanced loss function with a Generalized Cross Entropy loss function. Li, however, is directed to the analogous art of training a deep neural network for:
wherein the deep learning prediction model uses, as an objective function, a function obtained by combining a Class Balanced (OB) loss function and a Generalized Cross Entropy (GCE) loss function;
( Li > [0009], [0018–0020], and [0048–0050]: teaches training a deep semantic segmentation model using an RSRLF robust loss objective that combines a Generalized Cross Entropy loss with adaptive class balancing; specifically, Equation 2 defines (
L
R
S
R
L
F
=
L
G
C
E
×
w
k
... where
L
G
C
E
) is the Generalized Cross Entropy loss function and
w
k
is an adaptive class balance weight; Li further teaches, in [0057-0058] and [0063], that the GCE component improves robustness to noisy labels, while the adaptive class-balance component corrects differences in class sample size, and that the two components jointly optimize the deep-learning model, thereby teaching an objective function obtained by combining a class-balanced loss component and a GCE loss function. )
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Wu [as modified by Gagnon] to use Li’s combined GCE and adaptive class-balanced loss objective, because Wu already trains a ResNet-based image-classification model using stochastic gradient descent and performs data augmentation to balance data from different categories, while Li teaches a compatible ResNet-based deep-learning training technique in which GCE reduces the adverse effects of noisy or incorrect labels and adaptive class balancing corrects unequal class-sample distributions. Li further demonstrates that the combined objective improves classification performance compared with GCE alone. Accordingly, applying Li’s known robust loss objective to the Wu/Gagnon cataract-image model would have amounted to the predictable use of a known model-training technique according to its established function, with a reasonable expectation of improving training robustness and classification accuracy in the presence of class imbalance and erroneous labels.
Claim 37 is rejected under 35 U.S.C. §103 as being unpatentable over Wu [as modified by Gagnon and Li] in view of Jiang (Jiang et al, "Improving the Generalizability of Infantile Cataracts Detection via Lens Partition Strategies Based on Faster R-CNN", 2021).
Regarding claim 37, Wu [as modified by Gagnon and Li] teaches the method for diagnosis of cataract using deep learning of claim 36, further comprising:
Wu [as modified by Gagnon and Li] teaches acquiring multiple image sets under different slit-lamp illumination patterns, including retroillumination, and input into an ML model to output diagnostic data from acquired images, however Wu [as modified by Gagnon and Li] fails to teach a preprocessing step that crops or extracts only the lens region from both the slit-lamp and retroillumination images before classification where Jiang teaches:
before the step of input, a step of preprocessing in which a preprocessing unit extracts only a region corresponding to a lens from the slit lamp microscopic examination result image or the retroillumination examination result image using Faster R-CNN.
( Jiang, in [“Overall Architecture of LPS and Infantile Cataracts Diagnosis”], teaches that the overall architecture includes an automatic lens partition strategy (LPS) module and a diagnosis module, that two LPSs are present to localize the lens regions on the original slit-lamp images, and that the partitioned lens regions are input into a 50-layer residual CNN (ResNet50) for screening and three-degree grading of infantile cataracts; Jiang further teaches that an alternative partition strategy based on the Faster R-CNN method is used for detecting lens regions, that the RoI pooling operates on the proposed regions, and that two stages of Faster R-CNN are applied to predict the boundary coordinates of the lens region; Jiang also teaches that the input of the ResNet is the lens region partitioned by Faster R-CNN or Hough transform. In [Figure 3], which show the automatic lens partition results of Faster R-CNN, with the bold red rectangles denoting the boundaries of the partitioned lens. )
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claim invention, to modify Wu [as modified by Gagnon and Li]’s automated slit-lamp / retroillumination image-acquisition system with deep-learning cataract diagnosis workflow, to further include Jiang’s Faster R-CNN lens partition preprocessing, because Jiang teaches that localizing and partitioning the lens region before classification improves cataract screening and grading by focusing the downstream CNN on the diagnostically relevant lens area rather than irrelevant surrounding ocular structures. Incorporating Jiang’s known Faster R-CNN lens-region extraction into the Wu [as modified by Gagnon]'s cataract workflow would have been a predictable use of a known preprocessing technique for its established purpose, yielding the expected result of improved cataract classification from the acquired slit-lamp or retroillumination images.
Claims 38-40 are rejected under 35 U.S.C. §103 as being unpatentable over Wu [as modified by Gagnon and Li] in view of Chylack (Chylack et al, "The Lens Opacities Classification System III", 1993).
Regarding claim 38, Wu [as modified by Gagnon and Li] teaches the method for diagnosis of cataract using deep learning of claim 36,
Wu [as modified by Gagnon and Li] teaches using slit-lamp and retroillumination images in an ML-capable cataract workflow, and determining whether cataract is present, and severity / management output. However, Wu [as modified by Gagnon and Li] fails to teach the specific cataract-grading taxonomy and decision rule where Chylack teaches:
wherein the degree of cataract progression in a lens nucleus is determined by using an NO grade in which grades are classified based on the degree of opacity of a lens nucleus and an NC grade in which grades are classified based on the degree of browning of the lens nucleus,
( Chylack, in section [“Rules for Grading with the LOCS III System”, providing grading guidelines for evaluating slit-lamp and retroillumination images of cataracts: grading nuclear color (NC) and nuclear opalescence (NO), and in sub-sections [“To Grade NC”] and [“To Grade NO”], teaches grading nuclear color and nuclear opalescence by comparing the lens to NC and NO standards that represent increasing degrees of nuclear color (brunescence / browning) and opacity; see also [Fig. 5], showing NO1–NO6 and NC1–NC6 standards, at multiple time points / changes in the grades over time correspond to progression of lens nucleus. )
wherein the degree of cataract progression in the cortex of a lens is determined by using a CO grade in which grades are classified based on whether opaque opacity exists in a lens cortex portion of the retroillumination examination result image and on an area occupied by the opaque opacity in an entire lens, and
( Chylack, in sub-section [“To Grade C”], teaches that cortical opacity is graded by comparing the retroillumination image to the C standards and estimating the percentage of the pupillary area occupied by cortical opacity (CO); see also [Fig. 5], showing C1–C5 standards, at multiple time points / changes in the grades over time correspond to progression of lens cortex. )
wherein the degree of cataract progression in the posterior subcapsular cataract is determined by using a PSC grade in which grades are classified based on whether opaque opacity exists in a posterior part of the lens of the retroillumination examination result image and on an area occupied by the opaque opacity in the entire lens, and
( Chylack, in sub-section [“To Grade P”], teaches that posterior subcapsular cataract (PSC) is graded by comparing the retroillumination image to the P standards and estimating the percentage of the pupillary area occupied by posterior subcapsular opacity; see also [Fig. 5], showing P1–P5 standards, at multiple time points / changes in the grades over time correspond to progression of PSC. )
wherein the step of diagnosing cataract further includes a step of determining that the subject has no cataract when all of the NO grade, the NC grade, the CO grade and the PSC grade are 0; and a step of determining that the subject has cataract when the NO grade, the NC grade, the CO grade or PSC grade is not 0.
( Chylack, in [“Materials and Methods"], describing the LOCS II, and notes that LOCS II scores are “NO = 0 through 4; NC = 0–2 …” etc., thereby preserving a grade 0 corresponding to a normal lens for each opacity type. In light of these teaching that 0 denotes no cataract for each of NO, NC, C, and P, it would have been obvious to determine that a subject has no cataract when all four grades are 0, and that the subject has cataract when any of the NO, NC, CO, or PSC grades is non-zero. )
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claim invention, to modify Wu [as modified by Gagnon and Li] to use Chylack’s LOCS-based nuclear, cortical, and posterior subcapsular grading scheme because Wu [as modified by Gagnon and Li] already processes the same slit-lamp and retroillumination image types for cataract diagnosis and severity management, and Chylack provides the known standardized criteria for grading those cataract features from those image types. Using Chylack’s NO, NC, CO, PSC grading taxonomy in the Wu [as modified by Gagnon and Li]'s system would have been a predictable application of a known clinical grading standard to improve consistency and interpretability of the cataract determination and severity output, including the routine implementation that no graded opacity indicates no cataract and any non-zero graded opacity indicates cataract.
Regarding claim 39, Wu [as modified by Gagnon, Li and Chylack] teaches the method for diagnosis of cataract using deep learning of claim 38, wherein the cataract severity is divided into non-cataract, mild, moderate, and severe cataract levels, and
( Wu, in section [“Cataract diagnosis and management model for the AI agent” and “Evaluation Metrics,” teaches training a convolutional neural network to detect the presence of cataract and to predict cataract severity as a multiclass classification using LOCS-based NO, NC, C, and P grades as inputs and then dividing lens grading into severity-based grading categories: "A cataract with nuclear grading III or IV was defined as severe cataract or ‘referral’; otherwise, it was defined as mild cataract. If the primary evaluation decision was mild cataract with nuclear grading I~II". In view of these teachings, it would have been obvious to a person of ordinary skill in the art to implement overall cataract severity categories such as non-cataract, mild, moderate, and severe in the deep learning based diagnostic method. )
wherein the step of evaluating cataract severity further includes a step of evaluating the degree of cataract severity based on a greater value of the NO grade, the NC grade, the CO grade and the PSC grade,
( Chylack, in “Rules for Grading with the LOCS III System” and the subsections “To Grade NO,” “To Grade NC,” “To Grade C,” and “To Grade P”, teaches cataract severity grade for each opacity type, with higher values indicating greater severity of that particular opacity. LOCS–based method further teach using LOCS scores and cut-off values (for example, a cortical C score > 2 or high NO/NC values) as indicators of more advanced cataract and as potential thresholds for recommending cataract surgery; and [Fig. 5] shows the corresponding standards NO1-NO6, NC1-NC6, C1-C5, and P1-P5. )
and a step of extracting a visual acuity of the subject from the medical examination result and determining whether or not the visual acuity is equal to or greater than a certain standard.
( Wu, in page 1556 with [Fig. 5] and associated text, teaches a referral workflow in which visual acuity (VA) and a brief case history are collected, suspicious cases are referred for community healthcare evaluation, and the cataract AI agent uses the collected information in determining referral / follow-up management by considering the diagnosis and referable conditions. )
Regarding claim 40, Wu [as modified by Gagnon, Li and Chylack] teaches the method for diagnosis of cataract using deep learning of claim 39, wherein the step of providing treatment plan further includes:
in a case where the cataract severity is evaluated as non-cataract, a step of providing a schedule for a next examination when the visual acuity of the subject is equal to or greater than a certain standard, and outputting a phrase recommending a visit to a hospital when the visual acuity of the subject is less than a certain standard;
in a case where the cataract severity is evaluated as mild, a step of outputting a phrase recommending a visit to a hospital;
in a case where the cataract severity is evaluated as moderate, a step of outputting a phrase recommending a visit to a hospital when the visual acuity of the subject is equal to or greater than a certain standard, and outputting a phrase recommending surgery when the visual acuity of the subject is less than a certain standard; and
in a case where the cataract severity is evaluated as severe, a step of outputting a phrase recommending surgery.
( Wu, in section ["Results"] with [Fig. 5], in the cataract AI referral workflow, teaches collecting visual acuity (VA) and a brief case history, using the AI agent to determine cataract diagnosis and referable status, and applying a referral pattern / follow-up pathway based on the evaluation: At level II, suspicious cases based on self-monitoring are referred to community-based healthcare facilities. )
In view of Wu’s teachings, it would have been obvious to implement, within the Wu [as modified by Gagnon, Li and Chylack] system, a graduated treatment-plan output in which non-cataract cases with acceptable VA are scheduled for follow-up, non-cataract cases with reduced VA are directed for hospital evaluation, mild cataract cases are referred for hospital evaluation, moderate cataract cases are referred or escalated to surgery depending on whether VA is above or below a selected threshold, and severe cataract cases are directed to surgery, as a routine triage application of the taught cataract severity grading and VA-based referral / management framework.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEN KUDO whose telephone number is (571)272-4498. The examiner can normally be reached M-F 8am - 5pm.
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, Vincent Rudolph can be reached at 571-272-8243. 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.
KEN KUDO
Examiner
Art Unit 2671
/KEN KUDO/Examiner, Art Unit 2671
/VINCENT RUDOLPH/Supervisory Patent Examiner, Art Unit 2671
Prosecution Insights