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
Receipt is acknowledged of claim amendments with associated arguments/remarks, received May 21, 2026. Claims 1-10 are pending with amendments to claims 1, 7, 9.
Response to Arguments
Applicant’s arguments, see Remarks, pg 5, filed 05/21/2026, with respect to the objections of claim 1, 7, 9 for minor informalities has been fully considered and, in light of the associated amendments, is persuasive. Therefore, the objection has been withdrawn.
Applicant’s arguments, see pg 5-6, filed 05/21/2026, with respect to the rejections of claim 1-10 under 35 U.S.C. § 103 has been fully considered but is not persuasive.
Applicant argues the cited prior art Tosun et al (US 2020/0294231) does not teach or suggest the limitation "predict[ing] a subclass to which the cell belongs among a first benign subclass group and a first malignant subclass group" in claim 1 (Remarks – 05/21/2026, pg 5). Applicant argues classification/subclassification of cells would not include “Tosun's disclosure of "structures, inflammatory filtrates, locations in the WSI RO" as the subclasses, but this is incorrect since these are not subclasses for classifying cells” (Remarks – 05/21/2026, pg 5).
Respectfully, the examiner is not persuaded.
Claim 1 recites to “predict a subclass to which the cell belongs” but does not claim limitations to what is or not included in the “plurality of subclasses.” The specification states “The number of subclasses into which benign cells and malignant cells are classified is not limited. It is only necessary to employ a configuration in which benign cells and malignant cells are classified into the number of subclasses in which the first trained model can classify benign cells and malignant cells. (¶ [0019] with emphasis). The applicant did not provide remarks as to what subclasses are acceptable subclasses for the malignant or benign cells nor does the independent claim recite what subclasses are considered as the appropriate subclasses.
The broadest reasonable interpretation for a claim would therefore indicate any classification is acceptable for organizing and describing a subclass of the class. Tosun describes cells to be classified based on benign or malignant and further classifies the given class of cells based on phenotype features of the cells to categorize the cells based on risk (in cancerous cells, “ROIs may be sorted from benign to malignant, or if cancer is not present, then from benign to atypical.” (¶ [0096]) included in the analysis and citation of the Non-Final Rejection). The analysis is performed by performing simple image analysis (level 1), pointwise mutual information (level 2) and diagnostic labels (level 3) of the cells (¶ [0091]-[0092]) to classify the cells as normal/cancerous and further classified as benign/atypical (normal) and benign/malignant (cancerous) (¶ [0094]-[0096]) described in the analysis of the office action (“resulting in a multi-level classification (example of benign and high or low risk”) and further included citation pertaining to the machine learning training cited in the Non-Final Rejection.
Respectfully, the argument is not persuasive.
Regarding Claim 7, applicant relies on the argument pertaining to Tosun (as discussed above) and for the aforementioned reasons is not persuasive.
No further argument is presented.
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.
Claims 1-6, 8-10 are rejected under 35 U.S.C. 103 as being unpatentable over Teramoto et al (Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network, cited in Non-Final Rejection – 03/11/2026) in view of Tosun et al (US 2020/0294231, cited in Non-Final Rejection – 03/11/2026).
Regarding Claim 1, Teramoto et al teach a classification apparatus for classifying, for pathological diagnosis, a specimen cell as a benign cell or a malignant cell, said classification apparatus (DNCNN for classification of cellular patch images to classify cell as benign or malignant; Fig 3, 6 and 2.4 DCNN architecture, 2.6 Classification approaches) comprising:
at least one processor (DCNN implement on computer with XEON CPU and NVIDIA Quadro P5000 GPU; 2.7 Evaluation ¶ 2), the at least one processor carrying out:
an acquisition process of acquiring an image which includes the specimen cell as a subject (lung cells were collected by interventional cytology techniques and prepared for digital images, acquired with a microscope and still camera; Fig 1 and 2.1 Materials); and
a classification process of inputting the image which has been acquired by the acquisition means in the acquisition process into a first trained model (the image dataset are collected and input to a trained DCNN (trained with patch images) for classification of the cells as either benign or malignant; Fig 2-4, 6 and 2.4 DCNN architecture, 2.6 Classification approaches) and classifying the specimen cell as a benign cell or a malignant cell based on a result of prediction by the first trained model (cells are classified in the DCNN as benign or malignant based on 3 classification approaches in the DCNN; Fig 3, 6 and 2.6 Classification approaches),
the first trained model having been trained, while using as input an image that includes a cell as a subject (the DCNN is trained with patch images for classification of cells as either benign or malignant; Fig 2-4, 6 and 2.4 DCNN architecture, 2.6 Classification approaches).
Teramoto et al does not teach the first trained model having been trained to predict a subclass to which the cell belongs among a first benign subclass group in which benign cells are classified into a plurality of subclasses and a first malignant subclass group in which malignant cells are classified into a plurality of subclasses.
Tosun et al is analogous art pertinent to the technological problem addressed in the current application and teaches the first trained model (“trained model” for all claims is given its BRI as a general trained classification algorithm based on applicant’s drawing Fig 3 of classification section 12 shown as a single element with 3 sub-elements 121, 122, 123 and broad specification ¶ [0039], [0050]) having been trained to predict a subclass to which the cell belongs among a first benign subclass group in which benign cells are classified into a plurality of subclasses and a first malignant subclass group in which malignant cells are classified into a plurality of subclasses (whole slide images (WSI) 402 are analyzed to identify and classify the image based on a first-level classification for the cell basic characteristics (phenotype subclasses such as nuclear size/shape, mitosis counting, color etc) at a base level (first trained model) 404, which may then be used to classify cell type (such as benign or malignant) at a second level and a third classifying level (structures, inflammatory infiltrates, locations in the WSI ROI), resulting in a multi-level (Fig 3) classification (example of benign and high or low risk, with specific tissue structures 406, performed using machine learning, with supervised training (ground truth data labeling, Fig 8) of multiple AI algorithm models for given tissue (including lung ¶ [0171]); Fig 3-8 and ¶ [0091]-[0096], [0100]-[0105], [0110]-[0113]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Teramoto et al with Tosun et al including the first trained model having been trained to predict a subclass to which the cell belongs among a first benign subclass group in which benign cells are classified into a plurality of subclasses and a first malignant subclass group in which malignant cells are classified into a plurality of subclasses. By using deep learning classification of pathological samples with multiple levels of classification, samples may be classified with improved consistency in analysis and reduced computational time, resulting in improved patient safety, response time to diagnose, and improved diagnostic techniques, as recognized by Tosun et al (¶ [0005], [0013]-[0014]).
Regarding Claim 2, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 1 (as described above), wherein: an arbitrary subclass included in the first benign subclass group differs in visual sign or tissue type from the other subclasses included in the first benign subclass group (Tosun et al, the benign classified samples are further classified based on visual features into high-risk or low-risk; Fig 5, 7 and ¶ [0017]-[0018], [0096], [0110]-[0112]); and an arbitrary subclass included in the first malignant subclass group differs in visual sign or tissue type from the other subclasses included in the first malignant subclass group (Tosun et al, the malignant classified samples are further classified based on visual features into a risk category, based on severity of identified feature; Fig 5, 7 and ¶ [0095]-[0096], [0110]-[0112]).
Regarding Claim 3, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 1 (as described above), wherein: in the classification process, the at least one processor further inputs the image which has been acquired in the acquisition process into a second trained model which has been trained , while using as input an image that includes a cell as a subject, so as to predict a subclass to which the cell belongs among a second benign subclass group in which benign cells are classified into a plurality of subclasses and a second malignant subclass group in which malignant cells are classified into a plurality of subclasses (Tosun et al, whole slide images (WSI) are received 402 from acquired medical images and input to the trained DCNN model executed by a processor (¶ [0037]) and classify basic quantification (subclasses) at the equivalent first trained model (phenotype subclasses such as nuclear size/shape, mitosis counting, color etc) and identify and classify the cell type within the tissue structure at the equivalent second trained model to classify cells as benign or malignant (as associated with the basic quantitation) 404, with the model trained using supervised learning; Fig 3-8 and ¶ [0091]-[0096], [0100]-[0105], [0110]-[0113]); in the classification process, the at least one processor classifies the specimen cell as a benign cell or a malignant cell based further on a result of prediction by the second trained model (Tosun et al, cells are classified as benign or malignant based on the phenotype analysis combined with pointwise mutual information (PMI, ¶ [0088]); Fig 3-8 and ¶ [0091]-[0096], [0110]-[0113]); and the first benign subclass group differs from the second benign subclass group and/or the first malignant subclass group differs from the second malignant subclass group (Tosun et al, the cells classified as benign or malignant are classified based on phenotype profile; Fig 3-8 and ¶ [0091]-[0096], [0100]-[0105], [0110]-[0113]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Teramoto et al with Tosun et al including wherein: in the classification process, the at least one processor further inputs the image which has been acquired in the acquisition process into a second trained model which has been trained , while using as input an image that includes a cell as a subject, so as to predict a subclass to which the cell belongs among a second benign subclass group in which benign cells are classified into a plurality of subclasses and a second malignant subclass group in which malignant cells are classified into a plurality of subclasses; in the classification process, the at least one processor classifies the specimen cell as a benign cell or a malignant cell based further on a result of prediction by the second trained model; and the first benign subclass group differs from the second benign subclass group and/or the first malignant subclass group differs from the second malignant subclass group. By acquiring pathological samples from patient samples and analyzing samples using machine learning algorithms performing multiple levels of classification, samples may be classified with improved consistency in analysis and reduced computational time, resulting in improved patient safety, response time to diagnose, and improved diagnostic techniques, as recognized by Tosun et al (¶ [0005], [0013]-[0014]).
Regarding Claim 4, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 1 (as described above), wherein: in the classification process, the at least one processor further inputs the image which has been acquired in the acquisition process into a third trained model which has been trained, while using as input an image that includes a cell as a subject, so as to predict to which one of a benign cell and a malignant cell the cell belongs (Tosun et al, whole slide images (WSI) are received 402 from acquired medical images and input to the trained DCNN model executed by a processor (¶ [0037]) and classify basic quantification (subclasses) at equivalent first trained model (phenotype subclasses such as nuclear size/shape, mitosis counting, color etc), equivalent second trained model (identify and classify the cell type within the tissue structure as benign or malignant), and equivalent third trained model to identify spatial relationships and classify risk profile of classified cell (such as high-risk or low-risk) 404, with the model trained using supervised learning; Fig 3-8 and ¶ [0091]-[0096], [0100]-[0105], [0110]-[0113]); and in the classification process, the at least one processor classifies the specimen cell as a benign cell or a malignant cell based further on a result of prediction by the third trained model (Tosun et al, the benign or malignant classified cells are classified further based on risk, such as a high-risk or low-risk; Fig 3-8 and ¶ [0091]-[0096], [0100]-[0105], [0110]-[0113]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Teramoto et al with Tosun et al including wherein: in the classification process, the at least one processor further inputs the image which has been acquired in the acquisition process into a third trained model which has been trained, while using as input an image that includes a cell as a subject, so as to predict to which one of a benign cell and a malignant cell the cell belongs; and in the classification process, the at least one processor classifies the specimen cell as a benign cell or a malignant cell based further on a result of prediction by the third trained model. By acquiring pathological samples from patient samples and analyzing samples using machine learning algorithms performing multiple levels of classification, samples may be classified with improved consistency in analysis and reduced computational time, resulting in improved patient safety, response time to diagnose, and improved diagnostic techniques, as recognized by Tosun et al (¶ [0005], [0013]-[0014]).
Regarding Claim 5, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 4 (as described above), wherein: in a case where at least one of the results output from the respective plurality of trained models indicates that the specimen cell has been classified into the benign subclass group or as a benign cell, in the classification process, the at least one processor classifies the specimen cell as a benign cell (Tosun et al, cell images acquired medical images 402 and input to the trained DCNN model executed by a processor (¶ [0037]) are classified 404 as benign based on the phenotype combined with PMI analysis and further profiled and classified regarding a risk metric 406; Fig 3-4 and ¶ [0094]-[0096]).
Regarding Claim 6, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 1 (as described above), wherein: in the acquisition process, the at least one processor further acquires training data including a set of an image which includes a cell as a subject and a subclass to which the cell belongs (Tosun et al, the acquired medical images may be images used for a training dataset which are labeled by a user (pathologist) and input to the DCNN model to train the model for pattern identification of the cell class and subclass; Fig 8 and ¶ [0032], [0104]-[0105], [0113]); and the at least one processor further carries out a training process of training the trained model with use of the training data which has been acquired in the acquisition process (Tosun et al, the DCNN model executed by a processor (¶ [0037]) is trained with the labelled datasets for classification of cells as either benign or malignant; Fig 3 and ¶ [0032], [0104]-[0105], [0113]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Teramoto et al with Tosun et al including in the acquisition process, the at least one processor further acquires training data including a set of an image which includes a cell as a subject and a subclass to which the cell belongs; and the at least one processor further carries out a training process of training the trained model with use of the training data which has been acquired in the acquisition process. By using training image data, collected for and analyzed by a pathologist, supervised training is performed, resulting in the xAI system learning patterns for efficient and accurate classification with reduced computational time, resulting in improved patient safety, response time to diagnose, and improved diagnostic techniques, as recognized by Tosun et al (¶ [0005], [0013]-[0014]).
Regarding Claim 8, Teramoto et al teach a classification method using a classification apparatus for classifying, for pathological diagnosis, a specimen cell as a benign cell or a malignant cell (method using DNCNN for classification of cellular patch images to classify cell as benign or malignant; Fig 3, 6 and 2.4 DCNN architecture, 2.6 Classification approaches), said classification method comprising: process steps identical to claim 1 (as described above).
Regarding Claim 9, Teramoto et al teach a non-transitory storage medium comprising:
a program for causing a computer (DCNN implement on Xeon CP, recognized to be configured with memory (see NPL literature describing Xeon CPU configuration with high bandwidth memory; 2.7 Evaluation ¶ 2; examiner notes Tosun et al also teaches the non-volatile memory storing the classifier program, executed on processor ¶ [0248]) to function as a classification apparatus for classifying, for pathological diagnosis, a specimen cell as a benign cell or a malignant cell (DNCNN for classification of cellular patch images to classify cell as benign or malignant; Fig 3, 6 and 2.4 DCNN architecture, 2.6 Classification approaches), the program causing the computer (DCNN implement on computer with XEON CPU and NVIDIA Quadro P5000 GPU; 2.7 Evaluation ¶ 2) to carry out: process steps identical to claim 1 (as described above).
Regarding Claim 10, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 1 (as described above), wherein: the first trained model is a machine learning model which has been optimized (Teramoto et al, the DCNN (machine learning model) for classification is optimized during training based on stochastic gradient descent algorithm; 2.4 DCNN architecture).
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Teramoto et al (Automated classification of benign and malignant cells from lung cytological images using deep convolutional neural network, cited in Non-Final Rejection – 03/11/2026) in view of Tosun et al (US 2020/0294231, cited in Non-Final Rejection – 03/11/2026) and Gao et al (WO 2021/226493, cited in Non-Final Rejection – 03/11/2026).
Regarding Claim 7, Teramoto et al in view of Tosun et al teach the classification apparatus according to claim 1 (as described above), including wherein: in the acquisition process, the at least one processor acquires, when microscopically observing a cell of respiratory organs (Teramoto et al, microscopic images of lung cells were collected by interventional cytology techniques and prepared for digital images, acquired with a microscope and still camera; Fig 1 and 2.1 Materials).
Teramoto et al in view of Tosun et al does not teach observing a cell of respiratory organs taken with used of an endoscope, an image captured by a camera attached to a microscope.
Gao et al is analogous art pertinent to the technological problem addressed in the current application and teaches observing a cell of respiratory organs taken with use of an endoscope, an image captured by a camera attached to a microscope (a hyperspectral imaging surgical endoscope is used to capture the pulmonary pathological image data of the lungs and the image data is analyzed for the classification of the cellular tissue as benign or cancerous using CNN techniques; Fig 1, 4, 5 and ¶ [0021]-[0022], [0044], [0083]-[0086]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the current application to combine the teachings of Teramoto et al in view of Tosun et al with Gao et al including observing a cell of respiratory organs taken with used of an endoscope, an image captured by a camera attached to a microscope. By using endoscopic techniques to capture the image data, then applying machine learning models to classify the cell data to identify tumor margins, real-time intraoperative tumor assessment may be performed in vivo, thereby improving operating times to improve surgical management of lung cancer, as recognized by Gao et al (¶ [0005], [0007]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Ramirez et al (US 2019/0228527, cited in Non-Final Rejection – 03/11/2026) teach a cascade classifier architecture used to perform two-level analysis with a first level model to provide an accurate identification of the cell (malignant or benign, described as normal or abnormal cells) and a second level model to provide a categorial classification based on the first level (sub-level classification).
Yi et al (US 2018/0263568, cited in Non-Final Rejection – 03/11/2026) teach a endoscope imaging system combined with a machine learning model to classify imaged cells as benign or malignant, including sub-classification of the cells.
THIS ACTION IS MADE FINAL. 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 KATHLEEN M BROUGHTON whose telephone number is (571)270-7380. The examiner can normally be reached Monday-Friday 8:00-5:00.
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, John Villecco can be reached at (571) 272-7319. 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.
/KATHLEEN M BROUGHTON/Primary Examiner, Art Unit 2661
Prosecution Insights