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 .
DETAILED ACTION
The United States Patent & Trademark Office appreciates the response filed for the current application that is submitted on 05/04/2026. The United States Patent & Trademark Office reviewed the following documents submitted and has made the following comments below.
Priority
Acknowledgement is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. TW112101588, filed on 01/13/2023.
Amendment
Applicant submitted amendments on date 05/04/2026. The Examiner acknowledges the amendment and has reviewed the claims accordingly.
Overview
Claims 1, 3, 6-9, 11, and 14-17 are pending in this application and have been considered below.
Claims 2, 4-5, 10, 12 and 13 have been cancelled.
Claims 1, 3, 6-9, 11, and 14-17 are rejected.
Applicant Arguments:
In regards to Argument 1, Applicant/s state/s that “Claims 1-7 and 9-16 stands rejected under 35 U.S.C. 112(b) for failing to point out and distinctly claim the subject matter. In response, claims 2, 4-5, 10, 12, and 13 are cancelled, and claims 1, 3, 7, 9, 11, 15 and 16 are amended. After amendment, pending claims 1, 3, 6-9, 11, 14, 15, 16 and 17 no longer recite the indefinite term of "and/or".” Therefore, the 35 U.S.C. 112(b) rejection should be withdrawn from the pending claims.
In regards to Argument 2, Applicant/s state/s “the emphasized steps are NOT taught by Mahmood or Heindl references.” Specifically referring to the newly amended claim 1: steps of “(d) framing the detected attribute on each of the processed images of step (c) to produce a plurality of framed images” and “(g) combining each of the extracted sub-images of step (e) and each of the segmented sub-images of step (f)”
In regards to Argument 3, Applicant/s state/s “no expectation of success may be established on claims 1, 6 and 8 in view of the combined teaching of Mahmood and Heindl references.”
In regards to Argument 4, Applicant/s state/s “Applicant respectfully traverses this rejection based on the reason that neither Mahmood nor Heindl references has taught or suggested the combining step (g) as recite in the newly amended claim 1.”
In regards to Argument 5, Applicant/s state/s “The Examiner then alleged in the OA that such overlap would be synonymous to the step of ‘combining’ as recited in pending claim 1, for claim 1 is silent as to how the images are combined. Applicant respectfully disagrees.”
In regards to Argument 6, Applicant/s state/s “neither Mahmood or Heindl references has taught or suggested combining a segmented image with a previous un-segmented image to produce a combined image as required in step (g) of pending claim 1, in which each segmented sub-image (i.e., a patch) is combined with each of the extracted sub-image (i.e., ROI) to produce a combined image.”
In regards to Argument 7, Applicant/s state/s “Applicant respectfully submits that it is a fact that the overlay is NOT synonymous to ‘combining’ the images as alleged by the Examiner, as the patch comprising overlay as taught by Heindl is a combination of pixels from a first patch and pixels from a second patch in the same image…the patch comprising overlay is NOT the combined image of step (g), in which each combined image is a combination of pixels from a segmented image (e.g., a patch or a sub-image of step (f) of pending claim 1) and pixels from an un-segmented image (e.g., ROI or an extracted sub-image of step (e) of pending claim 1).”
In regards to Argument 8, Applicant/s state/s “no expectation of success may be established against pending claims 1, 6, and 8 in view of the combined teaching of Mahmood or Heindl references.”
In regards to Argument 9, Applicant/s state/s “the Examiner alleged that Park et al teaches generating at least one bounding box prediction to provide a particular prediction of breast cancer. Thus, it would have been obvious for a skilled artisan to modify Mahmood by detecting the attribute in the processed mammography images and framing the breast cancer prediction with a bounding box, that is taught by Park to make the invention that provides richer and more precise training signals to the model and leads to increased lesion detection performance. …Applicant respectfully traverses this rejection. As argument set forth above to refute the combined teaching of Mahmood or Heindl references, neither Mahmood nor Heindl references has taught or suggested the combining image step (g) recited in pending claim 1, thus, even with the framing image step taught by Park et al., still, the combined teaching of all cited references will never give rise to the method of pending claims 3 and 5” and therefore, the 35 U.S.C. rejection on claims 3 and 5 should be withdrawn.
In regards to Argument 10, Applicant/s state/s “Claim 7 stand rejected under 35 U.S.C. 103 as being unpatentable over Mahmood et al in view of Heindl et al, in further view of Juluru et al. In the OA, the Examiner alleged that Juluru teaches each of the processed images is segmented by use of a U-net architecture. Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Mahmood in view of Heindl by incorporating the step of segmenting the mammography images using a U-net algorithm that is taught by Juluru to make the invention that efficiently and accurately performs segmentation of biomedical images, such as mammography images. … Applicant respectfully traverses this rejection. As argument set forth above to refute the combined teaching of Mahmood or Heindl references, neither Mahmood nor Heindl references has taught or suggested the combining image step (g) recited in pending claim 1, thus, even with the segmenting image by use of a U-net as taught by Juluru et al., still, the combined teaching of all cited references will never give rise to the method of pending claim 7” and therefore, the 35 U.S.C. rejection on claim 7 should be withdrawn.
In regards to Argument 11, Applicant/s state/s “Claims 9 and 16-17 stand rejected under 35 U.S.C. 103 as being unpatentable over Nabavi et al. (U.S. Patent Pub. No. US 2024/0428577 A1, hereafter referred to as Nabavi) in view of Heindl et al. In response, claim 9 is amended in a similar manner as that made to pending claim 1. Applicant respectfully traverses this rejection. Nabavi reference teaches an apparatus for identifying and treating an abnormality, and a method of identifying and treating an abnormality by use of the apparatus. The apparatus comprises first and second convolutional neural networks (CNNs) joined by a distance function. In essence, Nabavi teaches identify abnormality by comparing features extracted from current and previous image data of a patient, in which the two images (i.e., current and previous image) are respectively processed in first and second CNNs. However, like Heindl reference, Nabavi reference also fails to teach or suggest the combining image step (g) recited in pending claim 9, in which a segmented image (i.e., each segmented sub-image of step (f) of pending claim 9) is combined with a previous un-segmented image (i.e., each extracted sub-image of step (e) of pending claim 9) to produce a combined image, which is a text image exhibiting the attribute for the mammographic image.”
In regards to Argument 12, Applicant/s state/s “no expectation of success may be established on pending claim 9 in view of the combined teaching of Nabavi and Heindl references. Dependent claims 16-17 are unobvious by virtue of being dependent from a non-obvious base claim 9.”
In regards to Argument 13, Applicant/s state/s “Claims 10-14 stand rejected under 35 U.S.C. 103 as being unpatentable over Nabavi et al. in view of Heindl et al., in further view of Park et al. In response, claims 10, 12, and 13 are cancelled, and claims 11 and 14 are pending. As argument set forth above to refute rejection made to claim 9, neither Nabavi nor Heindl reference has taught or suggested the combining image step (g) recited in pending claim 9, in which a segmented image (i.e., each segmented sub-image of step (f) of pending claim 9) is combined with a previous un-segmented image (i.e., each extracted sub-image of step (e) of pending claim 9) to produce a combined image, which is a text image exhibiting the attribute for the mammographic image. Thus, even with the framing image step taught by Park et al., still, the combined teaching of all cited references will never give rise to the method of pending claims 11 and 14” and therefore, the U.S.C. 103 rejections made to claims 11 and 14 should be withdrawn.
In regards to Argument 14, Applicant/s state/s “Claim 15 stand rejected under 35 U.S.C. 103 as being unpatentable over Nabavi et al. in view of Heindl et al., in further view of Juluru et al. Applicant respectfully traverses this rejection. As argument set forth above to refute the combined teaching of Nabavi and Heindl references, neither Nabavi nor Heindl references has taught or suggested the combining image step (g) recited in pending claim 9, thus, even with the segmenting image by use of a U-net as taught by Juluru et al., still, the combined teaching of all cited references will never give rise to the method of pending claim 15” and therefore the 35 U.S.C. 103 rejection made to claim 15 should be withdrawn.
Examiner’s Responses
In response to Argument 1, see remarks, filed 05/04/2026, with respect to the 112(b) rejections to claims 1-7 and 9-16, have been fully considered and are persuasive, therefore, the Examiner has withdrawn the rejections for 35 USC § 112 to pending claims 1, 3, 6-9, 11, 14, 15, 16 and 17 in response to Applicants amendments.
In response to Argument 2, see remarks, filed 05/04/2026, regarding applicant’s argument that the references fail to show certain features of the invention, particularly steps (d) and (g) in claim 1, have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made
for Claim 1 under 35 U.S.C. 103 in view of Dhungel et al. (NPL “A deep learning approach for the analysis of masses in mammograms with minimal user intervention,” 2017, hereafter referred to as Dhungel) in view of Mahmood et al (NPL “Breast lesions classifications of mammographic images using a deep convolutional neural network-based approach,” 2022, hereafter referred to as Mahmood) in further view of Wu et al. (NPL “Layer-Wise Pre-Training Low-Rank NMF Model for Mammogram-Based Breast Tumor Classification,” 2019, hereafter referred to as Wu).
The Examiner finds that Dhungel teaches on the amended claim language “(d) framing the detected attribute on each of the processed images of step (c) to produce a plurality of framed images” and “(g) combining each of the extracted sub-images of step (e) and each of the segmented sub-images of step (f), thereby producing a plurality of combined images respectively exhibiting the attribute for each of the mammographic images.”
Specifically, regarding step (d) “framing the detected attribute”, Dhungel teaches generating a set of mass candidates, comprising their bounding boxes and rough segmentation masks for a mammogram (3.2.1. Mass ROI detection). The Examiner interprets “framing the detected attribute” to be synonymous to placing bounding boxes around the detected attribute(s) in the mammogram in light of Applicant’s specification, which states, “the visual representation of the framed image will include the lesions that have been detected, outlined with bounding boxes” (Applicant’s specification, Paragraph [0053]). Regarding step (g), “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram are represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” In light of Applicant’s specification, which states, “the term ‘combined image’ … refers to an image composed of a segmented image and a cropped image derived from a raw image for object detection (i.e., lesion detection in the present application). … the combined images used for training a machine learning model serve as ‘reference images’” (Applicant’s specification, Paragraph [0039]). Therefore, the Examiner interprets a mammogram annotated with a bounding box “cropped image” derived from mammogram and a segmented image “segmentation map” to be a combined image since the claim is silent to the definition of combined image. Details of the rejection are below.
In response to Argument 3, see remarks filed 05/04/2026, regarding Applicant’s argument that there is no expectation of success established on claims 1, 6 and 8 in view of the combined teaching of Mahmood and Heindl references have been have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made is made for Claims 1 and 6 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu and for Claim 8 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in view of Wu in further view of Jurulu et al. (U.S. Patent Pub. No. 2025/0272831, hereafter referred to as Jurulu.).
Specifically, regarding Claim 1, Dhungel teaches a deep learning method for the analysis of masses in mammograms with minimal user intervention (Abstract). Dhungel teaches detection, segmentation, and classification of breast masses for mammograms. Dhungel performs mass detection using mass candidate generation (identifying bounding boxes/ROIs) and a refinement step. Dhungel performs segmentation using deep structured learning. Finally, classification is performed by training a CNN (1. Introduction). Dhungel does not explicitly disclose performing preprocessing steps on mammograms or a complex-sparse matrix factorization method. Mahmood is relied upon specifically for the image preprocessing techniques performed on the mammograms such as noise reduction, histogram equalization (CLAHE), cropping, image flipping, etc. (Mammogram preprocessing). It would have been obvious to a person having ordinary skill in the art to combine the detection, segmentation, and classification methods of Dhungel by performing the preprocessing steps on the mammography images taught by Mahmood to improve computational quality, image smoothing, and noise mitigation (Mammogram preprocessing). In addition, by performing geometric transforms on mammographic images in the preprocessing stage, such as image flipping, an enlarged dataset may be created to mitigate overfitting, boost the model’s generalization, and improve efficiency (Data augmentation). Dhungel nor Mahmood disclose a complex-sparse factorization method. Further, Wu teaches a modified nonnegative matrix factorization model called LPML-LRNMR for use in performing image-based breast tumor classification (Abstract). The model is an improved method of NMF (nonnegative matrix factorization), and the obtained coefficient matrix has sparse features (2.5 Breast Tumor Classification Based on LPML-LRNMP and ISSRC). It would have been obvious to a person having ordinary skill in the art to combine the method of Dhungel in view of Mahmood with the improved nonnegative matrix factorization method taught by Wu to effectively extract features of mammograms to greatly improve readability of the original mammograms (1 Introduction). Additional details of the rejections for claims 1, 6, and 8 are below.
In response to Argument 4, see remarks filed 05/04/2026, regarding Applicant’s argument that “neither Mahmood nor Heindl references has taught or suggested the combining step (g) as recite in the newly amended claim 1” have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 1 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu.
Specifically, Regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” Details of the rejection are below.
In response to Argument 5, see remarks filed 05/04/2026, regarding Applicant’s argument that such overlap is not synonymous to the step of ‘combining’ as recited in pending claim 1 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 1 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu. Details of the rejection are below.
Specifically, Regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” Details of the rejection are below.
In response to Argument 6, see remarks filed 05/04/2026, regarding Applicant’s argument that “neither Mahmood or Heindl references has taught or suggested combining a segmented image with a previous un-segmented image to produce a combined image as required in step (g) of pending claim 1, in which each segmented sub-image (i.e., a patch) is combined with each of the extracted sub-image (i.e., ROI) to produce a combined image.” Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 1 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu.
Specifically, Regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” Details of the rejection are below.
In response to Argument 7, see remarks filed 05/04/2026, regarding Applicant’s argument that “the overlay is NOT synonymous to ‘combining’ the images as alleged by the Examiner, as the patch comprising overlay as taught by Heindl is a combination of pixels from a first patch and pixels from a second patch in the same image…the patch comprising overlay is NOT the combined image of step (g), in which each combined image is a combination of pixels from a segmented image (e.g., a patch or a sub-image of step (f) of pending claim 1) and pixels from an un-segmented image (e.g., ROI or an extracted sub-image of step (e) of pending claim 1).” Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 1 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu.
Specifically, Regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” Details of the rejection are below.
In response to Argument 8, see remarks filed 05/04/2026, regarding Applicant’s argument that no expectation of success may be established against pending claims 1, 6, and 8 in view of the combined teaching of Mahmood or Heindl references has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claims 1 and 6 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu and Claim 8 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in view of Wu in further view of Jurulu et al. (U.S. Patent Pub. No. 2025/0272831, hereafter referred to as Jurulu.).
See response to Argument 3 above for combined teachings of Dhungel in view of Mahmood in further view of Wu for claims 1 and 6. Regarding Claim 8, Jurulu teaches the claim language “wherein the subject is a human.” Specifically, Jurulu teaches the individual, patient, or subject is a human (Paragraph [0043]). Details of the rejection are below.
In response to Argument 9, see remarks filed 05/04/2026, regarding Applicant’s argument that “neither Mahmood nor Heindl references has taught or suggested the combining image step (g) recited in pending claim 1, thus, even with the framing image step taught by Park et al., still, the combined teaching of all cited references will never give rise to the method of pending claims 3 and 5 and therefore, the 35 U.S.C. rejection on claims 3 and 5 should be withdrawn.” The examiner asserts that claim 5 has been cancelled. Regarding the argument made for claim 3, it has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 3 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in view of Wu in further view of Rogers et al. (U.S. Patent No. 6,091,841, hereafter referred to as Rogers).
Specifically, as stated in the arguments above, regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image. In addition, Rogers teaches filling holes in the mask by means of majority operation. The majority operation is a morphological operation in which each pixel in a binary image is turned on if a majority of its neighboring pixels are on. If the pixel is already on, it is left on (Col. 6, lines 65-67 and Col. 7, lines 1-12).” Details of the rejection are below.
In response to Argument 10, see remarks filed 05/04/2026, regarding Applicant’s argument that “Juluru teaches each of the processed images is segmented by use of a U-net architecture. Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Mahmood in view of Heindl by incorporating the step of segmenting the mammography images using a U-net algorithm that is taught by Juluru to make the invention that efficiently and accurately performs segmentation of biomedical images, such as mammography images. … Applicant respectfully traverses this rejection. As argument set forth above to refute the combined teaching of Mahmood or Heindl references, neither Mahmood nor Heindl references has taught or suggested the combining image step (g) recited in pending claim 1, thus, even with the segmenting image by use of a U-net as taught by Juluru et al., still, the combined teaching of all cited references will never give rise to the method of pending claim 7 and therefore, the 35 U.S.C. rejection on claim 7 should be withdrawn” has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 1 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Wu.
Specifically, as stated in the arguments above, regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” Additionally, regarding the motivation to combine Dhungel in view of Mahmood (see details of combination in rejection below) with Jurulu for segmenting the mammography image using a U-net architecture, Dhungel suggests “try(ing) better segmentation models… such as the U-net, which have produced state-of-the-art segmentation results in several computer vision datasets” (9. Future work). Details of the rejection are below.
In response to Argument 11, see remarks filed 05/04/2026, regarding Applicant’s argument that “Nabavi reference teaches an apparatus for identifying and treating an abnormality, and a method of identifying and treating an abnormality by use of the apparatus. The apparatus comprises first and second convolutional neural networks (CNNs) joined by a distance function. In essence, Nabavi teaches identify abnormality by comparing features extracted from current and previous image data of a patient, in which the two images (i.e., current and previous image) are respectively processed in first and second CNNs. However, like Heindl reference, Nabavi reference also fails to teach or suggest the combining image step (g) recited in pending claim 9, in which a segmented image (i.e., each segmented sub-image of step (f) of pending claim 9) is combined with a previous un-segmented image (i.e., each extracted sub-image of step (e) of pending claim 9) to produce a combined image, which is a text image exhibiting the attribute for the mammographic image” has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claims 9 and 16-17 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Jurulu.
Specifically, as stated in the arguments above, regarding step (g), “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” The Examiner interprets the segmentation map taught by Dhnugel to be the “segmented image” and the image patch defined by the bounding box to be the “unsegmented image.” Under Broadest Reasonable Interpretation (BRI), the segmentation map (segmented image) and image patch defined by a bounding box (un-segmented image) are “combined” since the segmentation map is “within” the image patch defined by the bounding box. Both are shown on top of the mammographic image to identify the lesion. Additionally, the Examiner interprets the mammogram annotated with the bounding box and lesion segmentation map/outline to be a “text image” since the claim is silent to the meaning of “text image.” Further, the Examiner interprets the combined image exhibits the attribute for the mammographic image since it is identified with a bounding box (region of interest) and a segmentation map of the lesion. Details of the rejection are below.
In response to Argument 12, see remarks filed 05/04/2026, regarding Applicant’s argument that “no expectation of success may be established on pending claim 9 in view of the combined teaching of Nabavi and Heindl references. Dependent claims 16-17 are unobvious by virtue of being dependent from a non-obvious base claim 9” has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claims 9 and 16-17 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Jurulu.
Specifically, Dhungel teaches a deep learning method for the analysis of masses in mammograms with minimal user intervention (Abstract). Dhungel teaches detection, segmentation, and classification of breast masses for mammograms. Dhungel performs mass detection using mass candidate generation (identifying bounding boxes/ROIs) and a refinement step. Dhungel performs segmentation using deep structured learning. Finally, classification is performed by training a CNN (1. Introduction). Mahmood is relied upon specifically for the image preprocessing techniques performed on the mammograms such as noise reduction, histogram equalization (CLAHE), cropping, image flipping, etc (Mammogram preprocessing). It would have been obvious to a person having ordinary skill in the art to combine the detection, segmentation, and classification methods of Dhungel by performing the preprocessing steps on the mammography images taught by Mahmood to improve computational quality, image smoothing, and noise mitigation (Mammogram preprocessing). In addition, by performing geometric transforms on mammographic images in the preprocessing stage, such as image flipping, an enlarged dataset may be created to mitigate overfitting, boost the model’s generalization, and improve efficiency (Data augmentation). Lastly, regarding claim 9, Jurulu is relied on for teaching step (i) of pending claim 9, which states “providing an anti-cancer treatment to the subject based on the breast lesion determined in step (h).” Jurulu teaches a clinician examining a subject may use information gathered from mammography images for diagnosing and deciding which treatment (e.g., a radiation therapy, immunotherapy, chemotherapy, surgery) to administer to the subject. It would have been obvious to a person having ordinary skill in the art to combine the lesion detection method of Dhungel in view of Mahmood with the anti-cancer treatment provided to the patient with cancer taught by Jurulu to reduce breast cancer mortality and morbidity by detecting the breast cancer early through mammograms (Dhungel, 1. Introduction).
In response to Argument 13, see remarks filed 05/04/2026, regarding Applicant’s argument that “Claims 10-14 stand rejected under 35 U.S.C. 103 as being unpatentable over Nabavi et al. in view of Heindl et al., in further view of Park et al. In response, claims 10, 12, and 13 are cancelled, and claims 11 and 14 are pending. As argument set forth above to refute rejection made to claim 9, neither Nabavi nor Heindl reference has taught or suggested the combining image step (g) recited in pending claim 9, in which a segmented image (i.e., each segmented sub-image of step (f) of pending claim 9) is combined with a previous un-segmented image (i.e., each extracted sub-image of step (e) of pending claim 9) to produce a combined image, which is a text image exhibiting the attribute for the mammographic image. Thus, even with the framing image step taught by Park et al., still, the combined teaching of all cited references will never give rise to the method of pending claims 11 and 14” and therefore, the U.S.C. 103 rejections made to claims 11 and 14 should be withdrawn” has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 9 and 14 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Jurulu and a new ground(s) of rejection is made for Claim 11 under 35 U.S.C 103 in view of Dhungel in view of Mahmood in view of Jurulu in further view of Rogers.
See above response to Argument 12 above for explanation of pending claim 9 rejection. Regarding Claim 14, Mahmood reads on the claim language “wherein in step (c), the attribute on the processed image of step (b) is detected by use of an object detection algorithm.” Specifically, Mahmood teaches an object detection algorithm, in particular, Mahmood teaches using CNNs such as VGGNet19, InceptionV3, ResNet152V2, InceptionResNetV2, and EfficientNetB5 pre-trained over ImageNet to fine-tuned over mammogram images to detect the breast masses (Transfer learning). Regarding Claim 11, Rogers reads on the claim language “further comprising mask filtering the framed image of step (d) to eliminate any mistaken attribute detected in step (c).” Specifically, Rogers teaches filling holes in the mask by means of majority operation. The majority operation is a morphological operation in which each pixel in a binary image is turned on if a majority of its neighboring pixels are on. If the pixel is already on, it is left on (Col. 6, lines 65-67 and Col. 7, lines 1-12). The Examiner will explain details of the rejection below.
In response to Argument 14, see remarks filed 05/04/2026, regarding Applicant’s argument that “Claim 15 stand rejected under 35 U.S.C. 103 as being unpatentable over Nabavi et al. in view of Heindl et al., in further view of Juluru et al. Applicant respectfully traverses this rejection. As argument set forth above to refute the combined teaching of Nabavi and Heindl references, neither Nabavi nor Heindl references has taught or suggested the combining image step (g) recited in pending claim 9, thus, even with the segmenting image by use of a U-net as taught by Juluru et al., still, the combined teaching of all cited references will never give rise to the method of pending claim 15” and therefore the 35 U.S.C. 103 rejection made to claim 15 should be withdrawn” has been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made for Claim 9 and 15 under 35 U.S.C. 103 in view of Dhungel in view of Mahmood in further view of Jurulu.
See above response to Argument 12 above for explanation of new pending claim 9 rejection. As stated in the arguments above, regarding step (g) “combining,” Dhungel teaches generating an annotated dataset where the annotations for the masses for the mammogram is represented by a segmentation map of the mass within the image patch defined by the bounding box (3.1. Dataset). Also see Fig. 5. Fig. 5 depicts a mammogram with highlighted mass detection (step (e) – extracted “sub-image” (bounding box)) and the segmented image (mass segmentation) “combined” since the segmentation map is contained within the bounding box or “sub-image.” Additionally, regarding claim 15 and the motivation to combine Dhungel in view of Mahmood (see details of combination in rejection below) with Jurulu for segmenting the mammography image using a U-net architecture, Dhungel suggests “try(ing) better segmentation models… such as the U-net, which have produced state-of-the-art segmentation results in several computer vision datasets” (9. Future work). The Examiner will maintain prior art Mahmood and details of the rejection are below.
Claim Objections
Claim 9 is objected to because of the following informalities:
In Claim 9, step (h), “by processing the text image of step (e)” should read, “processing the text image of step (g).”
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claim 3 is rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends. Claim 3 states: “the method of claim 2, further comprising…”. Claim 2 has been cancelled and therefore, cannot be depended upon. Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
Claim Rejections - 35 USC § 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.
Claims 1 and 6 are rejected under 35 U.S.C. 103 as being unpatentable over Dhungel et al. (NPL “A deep learning approach for the analysis of masses in mammograms with minimal user intervention,” 2017, hereafter referred to as Dhungel) in view of Mahmood et al. (NPL “Breast lesions classifications of mammographic images using a deep convolutional neural network-based approach,” 2022, hereafter referred to as Mahmood) in further view of Wu et al. (NPL “Layer-Wise Pre-Training Low-Rank NMF Model for Mammogram-Based Breast Tumor Classification,” 2019, hereafter referred to as Wu).
Regarding Claim 1, Dhungel teaches a method for building a model for determining a breast lesion in a subject (Abstract, Dhungel teaches a method for detecting, segmenting and classifying breast masses from mammograms. For the detection stage, a cascade of deep learning methods is proposed. For the segmentation, deep structured output learning that is subsequently refined is used. Lastly, for the classification, a deep learning classifier is proposed.), comprising:(a) obtaining a plurality of mammographic images of the breast from the subject (1. Introduction, Fig. 1, Dhungel teaches the detection, segmentation and classification accuracy produced by the method are measured on the publicly available INbreast dataset which consists of 410 FFDM mammograms of the left and right breasts from 115 patients from two views.),
PNG
media_image1.png
280
410
media_image1.png
Greyscale
in which each of the mammographic images comprises an attribute of the breast lesion selected from the group consisting of location (Abstract, Dhungel teaches location of breast masses in mammograms.), margin (Fig. 1, 8. Discussion, Dhungel teaches contours. Additionally, the Examiner interprets the mammograms have edges/contours. See Fig, 1 above (outlined edges of lesions).), calcification (9. Future work, Dhengel teaches mammograms have micro-calcifications), lump (Fig. 1, Dhungel teaches mammograms with breast masses. The Examiner interprets a “mass” and a “lump” in a mammogram to be synonymous since they are synonymous terms in the art.), mass (Abstract, Dhungel teaches classifying breast masses in mammograms.), shape (1. Introduction, Dhungel teaches breast masses have a large variability in terms of shape.), size (1. Introduction, Dhungel teaches breast masses have a large variability in terms of size.) status of breast lesion, and a combination thereof (Abstract, Fig. 1, Dhungel teaches masses are classified as malignant or benign. Under Broadest Reasonable Interpretation (BRI), the Examiner interprets the claim language of “selected from” and the combination isn’t specified to be only one or a few of the listed limitations need to be met.); (3.2. Mass Detection, 3.2.1. Mass ROI detection, Fig. 3, Dhungel teaches a mass detection algorithm which consists of a cascade of classifiers, where the main goal of each stage is to keep the true positive detections and then improving the precision of the bounding box detection. The first stage of detection consists of the generation of a set of mass candidates comprising their bounding boxes and a rough segmentation outline for a mammogram.);
[AltContent: arrow]
PNG
media_image2.png
259
1495
media_image2.png
Greyscale
(d) framing the detected attribute on each of the processed images of step (c) to produce a plurality of framed images (3.2.1. Mass ROI detection, Dhungel teaches the detection stage consists of the generation of a set of mass candidates, comprising their bounding boxes. The Examiner interprets “framing the detected attribute” to include placing bounding boxes around the detected attribute in light of Applicant’s specification, which states “the visual representation of the framed image will include the lesions that have been detected and outlines with bounding boxes” (Applicant’s specification, Paragraph [0053]).); (e) cropping each of the framed images of step (d) to produce a plurality of extracted sub-images (3.2.1. Mass ROI detection, 4. Mass segmentation, Fig. 3, Dhungel teaches taking each bounding box and extracting an image patch from x, which is resized to MxM. Once each bounding box is estimated, it is used to crop the image patch to a low-resolution patch of size MxM.); (f) segmenting each of the plurality of extracted sub-images of step (e) to produce a plurality of segmented sub-images (4. Mass segmentation, Fig. 4, Dhungel teaches the mass segmentation algorithm uses deep structured output learning to produce a segmentation on a low-resolution input image patch. The segmentation map is estimated in the low-resolution image patch.);
PNG
media_image3.png
456
1518
media_image3.png
Greyscale
(g) combining each of the extracted sub-images of step (e) and each of the segmented sub-images of step (f), thereby producing a plurality of combined images respectively exhibiting the attribute for each of the mammographic images (3.1. Dataset, Fig. 5, Dhungel teaches the annotated dataset with annotations Ai for the masses for mammogram i represented by
PNG
media_image4.png
92
320
media_image4.png
Greyscale
; where
PNG
media_image5.png
55
320
media_image5.png
Greyscale
represents the left-top position (x,y) and the width w and height h of the bounding box of the jth mass of the ith mammogram,
PNG
media_image6.png
41
145
media_image6.png
Greyscale
{0,1} represents the segmentation map of the mass within the image patch defined by the bounding box
PNG
media_image7.png
45
55
media_image7.png
Greyscale
. The Examiner interprets the extracted sub-image (bounding box) and the segmented sub-image (segmentation map) are “combined” since the segmentation map is “within the image patch defined by the bounding box.” The plain meaning of “within” is defined as “inside the range of (an area or boundary” or “inside (something). The word “combined” is defined as “describing something that is formed, produced, or consisting of two or more things joined together.” The Examiner interprets that the image containing a segmentation map within a bounding box (sub-image) annotated on the mammogram is a combined image since the segmentation map is within the bounding box, and therefore is consisting of two things “joined together” in the same boundary.) and (h) classifying the plurality of combined images of step (g) (3.1. Dataset, 1. Introduction, Dhungel teaches denoting the class label of the mass that can either be benign (i.e., BI-RADS class 2, 3) or malignant (i.e., BI-RADS class 4, 5, 6). The classification stage, based on deep learning methods, takes the appearance and shape from the automatically detected and segmented bounding boxes and produces the final mass classification.) (Fig. 2 caption, Dhungel teaches training a CNN in two steps, where the first step is a regressor that estimates hand-crafted features followed by a second step that fine-tunes the model based on the mass classification problem.).
Dhungel does not explicity disclose (b) producing a plurality of processed images via subjecting each of the plurality of mammographic images to image treatments selected from the group consisting of image cropping, image denoising, image flipping, histogram equalization, image padding, and a combination thereof and with the aid of a complex-sparse matrix factorization method.
Mahmood is in the same field of art of using a convolutional neural network to classify the region of interest of breast masses. Further, Mahmood teaches (b) producing a plurality of processed images via subjecting each of the plurality of mammographic images to image treatments selected from the group consisting of image cropping (Data augmentation, Cropping, Fig. 2(f), Mahmood teaches creating an enlarged dataset by executing geometric transforms on the small dataset, such as cropping. Fig. 2f shows the cropped images.), image denoising (Results and analysis, Mammogram preprocessing, Mahmood teaches each image is preprocessed to eliminate noise.), image flipping (Data augmentation, Fig. 2(b-c) Mahmood teaches creating an enlarged dataset by executing geometric transforms on the small dataset, such as flipping.), histogram equalization (Mammogram preprocessing, Mahmood teaches using the contrast limited adaptive histogram equalization (CLAHE) method to enhance the overall quality of the mammogram images.), image padding, and a combination thereof. (Mammogram preprocessing, Mahmood teaches applying image enhancement approaches, such as edges. Under Broadest Reasonable Interpretation (BRI), the Examiner interprets the claim language of “selected from” and the combination isn’t specified to be only one or a few of the listed limitations need to be met.).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel by preprocessing the mammographic images that is taught by Mahmood, to make the invention that performs preprocessing steps on the image prior to the detection, segmentation, and classification method steps on the mammograms; thus, one of ordinary skilled in the art would be motivated to combine the references since in mammograms, a significant number of anomalies are misinterpreted due to poor quality images with artifacts, pectoral tissue, low visibility, and interference with noise, which impairs image quality, resulting in a high false-positive rate. Preprocessing of images improves computational quality, image smoothing, and noise mitigation (Mahmood, Mammogram preprocessing). Additionally, executing geometric transforms on a small dataset, such as image flipping, can create variation in the dataset to prevent overfitting, boost the model’s generalization and improve efficiency (Mahmood, Data augmentation).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
Although Dhungel in view of Mahmood discloses (h) classifying the plurality of combined images of step (g). Dhungel in view of Mahmood does not explicitly disclose (h) classifying the plurality of combined images of step (g) with the aid of a complex-sparse matrix factorization method.
Wu is in the same field of art of performing image-based tumor classification. Further, Wu teaches (h) classifying the plurality of combined images of step (g) with the aid of a complex-sparse matrix factorization method (Abstract, 2.1 Layer-Wise Pre-training Multilayer Low-Rank NMF Model, 2.5 Breast Tumor Classification Based on LPML-LRNMF and ISSRC, Fig. 2, Wu teaches a modified nonnegative matrix factorization (NMF) model called LPML-LRNMP. For completing classification, an inverse projection sparse representation model is introduced to exploit information embedded in existing samples. LPML-LRNMP is an improved method of NMF, and the obtained coefficient matrix has sparse features. A mammogram-based tumor classification scheme is proposed by integrating LPML-LRNMF feature representation learning and ISSRC.).
PNG
media_image8.png
719
1053
media_image8.png
Greyscale
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel in view of Mahmood by integrating a nonnegative matrix factorization method into the mammogram classification procedure that is taught by Wu, to make the invention that classifies the combined images (from Dhungel) with the aid of a nonnegative matrix factorization method; thus, one of ordinary skilled in the art would be motivated to combine the references since the nonnegative matrix factorization (NMF) does not pay attention to category information, and explores useful information contained in all available samples simultaneously, even if there are only a small number of training samples (Wu, 1 Introduction). In addition, Wu considers adding more targeted prior information to optimize the model (4 Conclusions).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
In regards to Claim 6, Dhungel in view of Mahmood in further view of Wu teaches the method of claim 1, wherein in step (c), the attribute on each of the processed images of step (b) is detected by use of an object detection algorithm (Transfer learning, Mahmood teaches the study uses CNNs such as VGGNet19, InceptionV3, ResNet152V2, InceptionResNetV2, and EfficientNetB5 pre-trained over ImageNet to fine-tuned over mammogram images to detect the breast masses).
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Dhungel et al. (NPL “A deep learning approach for the analysis of masses in mammograms with minimal user intervention,” 2017, hereafter referred to as Dhungel) in view of Mahmood et al (NPL “Breast lesions classifications of mammographic images using a deep convolutional neural network-based approach,” 2022, hereafter referred to as Mahmood) in view of Wu et al. (NPL “Layer-Wise Pre-Training Low-Rank NMF Model for Mammogram-Based Breast Tumor Classification,” 2019, hereafter referred to as Wu) in further view of Rogers et al. (U.S. Patent No. 6,091,841, hereafter referred to as Rogers).
In regards to Claim 3, Dhungel in view of Mahmood in further view of Wu discloses the method of claim 1 (**The Examiner is treating the claim as if it is dependent on claim 1. Claim 2 has been cancelled.**).
Dhungel in view of Mahmood in further view of Wu does not explicitly disclose further comprising mask filtering the framed image of step (d) to eliminate any mistaken attribute detected in step (c) (Col. 6, lines 65-67 and Col. 7, lines 1-12, Rogers teaches filling holes in the mask by means of majority operation. The majority operation is a morphological operation in which each pixel in a binary image is turned on if a majority of its neighboring pixels are on. If the pixel is already on, it is left on.).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel in view of Mahmood in further view of Wu by filling holes in the mask that is taught by Rogers, to make the invention that generates an accurate silhouette (binary mask) of the breast which is used in later processing steps to decrease the time required for processing the mammogram image; thus, one of ordinary skilled in the art would be motivated to combine the references since if the holes in the mask are not filled in, the multiple erode steps may break the mask into disjointed sections (Rogers, Col. 7, lines 4-12).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
Claims 7 and 8 are rejected under 35 U.S.C. 103 as being unpatentable over Dhungel et al. (NPL “A deep learning approach for the analysis of masses in mammograms with minimal user intervention,” 2017, hereafter referred to as Dhungel) in view of Mahmood et al. (NPL “Breast lesions classifications of mammographic images using a deep convolutional neural network-based approach,” 2022, hereafter referred to as Mahmood) in view of Wu et al. (NPL “Layer-Wise Pre-Training Low-Rank NMF Model for Mammogram-Based Breast Tumor Classification,” 2019, hereafter referred to as Wu) in further view of Jurulu et al. (U.S. Patent Pub. No. 2025/0272831, hereafter referred to as Jurulu).
Regarding Claim 7, Dhungel in view of Mahmood in further view of Wu discloses the method of claim 1.
Dhungel in view of Mahmood in further view of Wu does not explicitly disclose wherein in step (f), each of the plurality of extracted sub-images of step (e) is segmented by use of a U-net architecture.
Jurulu is in the same field of art of identifying regions of interest corresponding to a dense area (lesion) in mammograms. Further, Jurulu discloses wherein in step (f), each of the plurality of extracted sub-images of step (e) is segmented by use of a U-net architecture (Paragraphs [0034], [0109], Juluru teaches an automatic mammographic dense area (MDA) segmentation algorithm (measure of breast density) for the unaffected breast using a fully convolutional neural network (FCNN) approach based on the U-net architecture. The images were cropped during the pre-processing stage to the breast region (i.e., “sub-images).).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel in view of Mahmood in further view of Wu by segmenting the cropped images or “sub-images” using a U-net architecture that is taught by Jurulu, to make the invention that performs image segmentation on the cropped mammograms to estimate the dense regions or “lesions”; thus, one of ordinary skilled in the art would be motivated to combine the references since it would have been obvious to replace the segmentation method in Dhungel with a U-net architecture since Dhungel suggests “trying better segmentation models, such as … the U-net, which have produced state-of-the-art segmentation results in several computer vision datasets” (Dhungel, 9. Future Work).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
In regards to Claim 8, Dhungel in view of Mahmood in further view of Wu discloses the method of claim 1.
Dhungel in view of Mahmood in further view of Wu does not explicitly disclose wherein the subject is human.
Jurulu is in the same field of art of identifying regions of interest corresponding to a dense area (lesion) in mammograms. Further, Jurulu teaches wherein the subject is human (Paragraph [0043], Juluru teaches the individual, patient, or subject is a human.).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel in view of Mahmood in further view of Wu by performing steps (a)-(f) on human mammogram images that is taught by Jurulu, to make the invention that classifies malignant and benign attributes/lesions on human mammograms to detect breast cancer; thus, one of ordinary skilled in the art would be motivated to combine the references since breast screening using mammograms can reduce breast cancer mortality and morbidity in humans (Dhungel, 1. Introduction).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
Claims 9 and 14-17 are rejected under 35 U.S.C. 103 as being unpatentable over Dhungel et al. (NPL “A deep learning approach for the analysis of masses in mammograms with minimal user intervention,” 2017, hereafter referred to as Dhungel) in view of Mahmood et al, (NPL “Breast lesions classifications of mammographic images using a deep convolutional neural network-based approach,” 2022, hereafter referred to as Mahmood) in further view of Jurulu et al. (U.S. Patent Pub. No. 2025/0272831 A1, hereafter referred to as Jurulu).
Regarding Claim 9, Dhungel teaches (1. Introduction, Fig. 1, Dhungel teaches the detection, segmentation and classification accuracy produced by the method are measured on the publicly available INbreast dataset which consists of 410 FFDM mammograms of the left and right breasts from 115 patients from two views.), in which the mammographic image comprises an attribute of the breast lesion selected from the group consisting of location (Abstract, Dhungel teaches location of breast masses in mammograms.), margin (Fig. 1, 8. Discussion, Dhungel teaches contours. Additionally, the Examiner interprets the mammograms have edges/contours. See Fig, 1 above (outlined edges of lesions).), calcification (9. Future work, Dhengel teaches mammograms have micro-calcifications), lump (Fig. 1, Dhungel teaches mammograms with breast masses. The Examiner interprets a “mass” and a “lump” in a mammogram to be synonymous since they are synonymous terms in the art.), mass (Abstract, Dhungel teaches classifying breast masses in mammograms.), shape (1. Introduction, Dhungel teaches breast masses have a large variability in terms of shape.), size (1. Introduction, Dhungel teaches breast masses have a large variability in terms of size.) status of breast lesion, and a combination thereof (Abstract, Fig. 1, Dhungel teaches masses are classified as malignant or benign. Under Broadest Reasonable Interpretation (BRI), the Examiner interprets the claim language of “selected from” and the combination isn’t specified to be only one or a few of the listed limitations need to be met.); (3.2. Mass Detection, 3.2.1. Mass ROI detection, Fig. 3, Dhungel teaches a mass detection algorithm which consists of a cascade of classifiers, where the main goal of each stage is to keep the true positive detections and then improving the precision of the bounding box detection. The first stage of detection consists of the generation of a set of mass candidates comprising their bounding boxes and a rough segmentation outline for a mammogram.); (d) framing the detected attribute on each of the processed images of step (c) to produce a plurality of framed images (3.2.1. Mass ROI detection, Dhungel teaches the detection stage consists of the generation of a set of mass candidates, comprising their bounding boxes. The Examiner interprets “framing the detected attribute” to include placing bounding boxes around the detected attribute in light of Applicant’s specification, which states “the visual representation of the framed image will include the lesions that have been detected and outlines with bounding boxes” (Applicant’s specification, Paragraph [0053]).); (e) cropping each of the framed images of step (d) to produce a plurality of extracted sub-images (3.2.1. Mass ROI detection, 4. Mass segmentation, Fig. 3, Dhungel teaches taking each bounding box and extracting an image patch from x, which is resized to MxM. Once each bounding box is estimated, it is used to crop the image patch to a low-resolution patch of size MxM.); (f) segmenting the extracted sub-image of step (e) to produce a segmented sub-image (4. Mass segmentation, Fig. 4, Dhungel teaches the mass segmentation algorithm uses deep structured output learning to produce a segmentation on a low-resolution input image patch. The segmentation map is estimated in the low-resolution image patch.); (g) combining the extracted sub-image of step (e) and the segmented sub- image of step (f), thereby producing a text image exhibiting the attribute for the mammographic image (3.1. Dataset, Fig. 5, Dhungel teaches the annotated dataset with annotations Ai for the masses for mammogram i represented by
PNG
media_image4.png
92
320
media_image4.png
Greyscale
; where
PNG
media_image5.png
55
320
media_image5.png
Greyscale
represents the left-top position (x,y) and the width w and height h of the bounding box of the jth mass of the ith mammogram,
PNG
media_image6.png
41
145
media_image6.png
Greyscale
{0,1} represents the segmentation map of the mass within the image patch defined by the bounding box
PNG
media_image7.png
45
55
media_image7.png
Greyscale
. The Examiner interprets the extracted sub-image (bounding box) and the segmented sub-image (segmentation map) are “combined” since the segmentation map is “within the image patch defined by the bounding box.” The plain meaning of “within” is defined as “inside the range of (an area or boundary” or “inside (something). The word “combined” is defined as “describing something that is formed, produced, or consisting of two or more things joined together.” The Examiner interprets that the image containing a segmentation map within a bounding box (sub-image) annotated on the mammogram is a combined image since the segmentation map is within the bounding box, and therefore is consisting of two things “joined together” in the same boundary.); (h) determining the breast lesion of the subject by processing the text image of step (e) within the model established by the method of claim 1(Fig. 14, Dhungel teaches an accurate automatically generated ROI and segmentation as shown in Fig. 14. Then, mass classification is performed based on the model on features from CNN pre-training. The Examiner interprets “determining the breast lesion” to include classifying the breast lesion since the claim is silent to the meaning of “determining.” Additionally, as shown in Fig. 14 below, each mammogram is classified (benign or malignant) and assigned a BI-RADS score.);
PNG
media_image9.png
520
384
media_image9.png
Greyscale
Dhungel does not explicitly disclose a method for treating breast cancer via determining a breast lesion in a subject, (b) producing a processed image via subjecting the mammographic image to image treatments selected from the group consisting of image cropping, image denoising, image flipping, histogram equalization, image padding, and a combination thereof; and (i) providing an anti-cancer treatment to the subject based on the breast lesion determined in step (h).
Mahmood is in the same field of art of using a convolutional neural network to classify the region of interest of breast masses. Further, Mahmood teaches (b) producing a processed image via subjecting the mammographic image to image treatments selected from the group consisting of image cropping (Data augmentation, Cropping, Fig. 2(f), Mahmood teaches creating an enlarged dataset by executing geometric transforms on the small dataset, such as cropping. Fig. 2f shows the cropped images.), image denoising (Results and analysis, Mammogram preprocessing, Mahmood teaches each image is preprocessed to eliminate noise.), image flipping (Data augmentation, Fig. 2(b-c) Mahmood teaches creating an enlarged dataset by executing geometric transforms on the small dataset, such as flipping.), histogram equalization (Mammogram preprocessing, Mahmood teaches using the contrast limited adaptive histogram equalization (CLAHE) method to enhance the overall quality of the mammogram images.), image padding, and a combination thereof. (Mammogram preprocessing, Mahmood teaches applying image enhancement approaches, such as edges. Under Broadest Reasonable Interpretation (BRI), the Examiner interprets the claim language of “selected from” and the combination isn’t specified to be only one or a few of the listed limitations need to be met.).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel by preprocessing the mammographic images that is taught by Mahmood, to make the invention that performs preprocessing steps on the image prior to the detection, segmentation, and classification steps on the mammograms; thus, one of ordinary skilled in the art would be motivated to combine the references since in mammograms, a significant number of anomalies are misinterpreted due to poor quality images with artifacts, pectoral tissue, low visibility, and interference with noise, which impairs image quality, resulting in a high false-positive rate. Preprocessing of images improves computational quality, image smoothing, and noise mitigation (Mahmood, Mammogram preprocessing).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
Dhungel in view of Mahmood does not explicitly disclose a method for treating a-breast cancer via determining a breast lesion in a subject and (i) providing an anti-cancer treatment to the subject based on the breast lesion determined in step (h).
Juluru is in the same field of art of identifying regions of interest corresponding to a dense area (lesion) in mammograms. Further, Juluru teaches a method for treating a breast cancer via determining a breast lesion in a subject (Paragraph [0083], Fig. 5, Juluru teaches determining whether the subject should be administered with a treatment for breast cancer based on the density value determined from a mammogram. The treatment may include for example, radiation therapy, immunotherapy, chemotherapy or surgery, among others.) and (i) providing an anti-cancer treatment to the subject based on the breast lesion determined in step (h) (Paragraph [0085], Juluru teaches a clinician examining the subject may administer treatment (e.g., radiation therapy, immunotherapy, chemotherapy, surgery) based on the indication. The indication may be any one or more of the following: the identifier for the subject, the mammogram, the segmentation map, and the density value for the subject.).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel in view of Mahmood by providing an anti-cancer treatment to the patient based on the identification of a cancerous lesion in the mammogram that is taught by Jurulu, to make the invention that utilizes the information gathered from processing the mammographic image to make an informed decision for which treatment method to administer to the patient based on density, etc.; thus, one of ordinary skilled in the art would be motivated to combine the references since breast cancer is curable if detected early, increasing likelihood of patient survival. In addition, timely detection of breast cancer masses is imperative for proper medication due to their modest size when patients exhibit no initial symptoms (Introduction, Mahmood).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
In regards to Claim 14, Dhungel in view of Mahmood in further view of Jurulu discloses the method of claim 9, wherein in step (c), the attribute on the processed image of step (b) is detected by use of an object detection algorithm (Transfer learning, Mahmood teaches the study uses CNNs such as VGGNet19, InceptionV3, ResNet152V2, InceptionResNetV2, and EfficientNetB5 pre-trained over ImageNet to fine-tuned over mammogram images to detect the breast masses).
In regards to Claim 15, Dhungel in view of Mahmood in further view of Jurulu discloses the method of claim 9, wherein in step (f), each of the extracted sub-image of step (e) is segmented by use of a U-net architecture (Paragraphs [0034], [0109], Juluru teaches an automatic mammographic dense area (MDA) segmentation algorithm (measure of breast density) for the unaffected breast using a fully convolutional neural network (FCNN) approach based on the U-net architecture. The images were cropped during the pre-processing stage to the breast region. The Examiner interprets a cropped image to be a sub-image, which is then segmented by the U-net architecture).
In regards to Claim 16, Dhungel in view of Mahmood in further view of Jurulu discloses the method of claim 9, wherein in step (i), the anti-cancer treatment is selected from the group consisting of a surgery (Paragraphs [0009], [0014], Juluru teaches administering surgery.), a radiofrequency ablation (Paragraph [0014], Juluru teaches administering radiation therapy to the subject.), a systemic chemotherapy (Paragraphs [0009], [0014], Juluru teaches administration of chemotherapy to the subject.), a transarterial chemoembolization (TACE), an immunotherapy (Paragraph [0014], Juluru teaches administration of immunotherapy to the subject.), a targeted drug therapy, a hormone therapy, and a combination thereof (Under Broadest Reasonable Interpretation, the Examiner interprets the claim language of “selected from the group consisting of” to be only one or a few of the limitations need to be met.).
In regards to Claim 17, Dhungel in view of Mahmood in further view of Jurulu the method of claim 9, wherein the subject is a human (Paragraph [0043], Juluru teaches the individual, patient, or subject is a human.).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Dhungel et al. (NPL “A deep learning approach for the analysis of masses in mammograms with minimal user intervention,” 2017, hereafter referred to as Dhungel) in view of Mahmood et al. (NPL “Breast lesions classifications of mammographic images using a deep convolutional neural network-based approach,” 2022, hereafter referred to as Mahmood) in view of Jurulu et al. (U.S. Patent Pub. No. 2025/0272831 A1, hereafter referred to as Jurulu) in further view of Rogers et al. (U.S. Patent No. 6,091,841, hereafter referred to as Rogers).
Regarding Claim 11, Dhungel in view of Mahmood in further view of Jurulu disclose the method of claim 9.
Dhungel in view of Mahmood in further view of Jurulu do not explicitly disclose mask filtering the framed image of step (d) to eliminate any mistaken attribute detected in step (c).
Rogers is in the same field of art of detecting suspicious regions, such as microcalcifications in mammograms. Further, Rogers teaches mask filtering the framed image of step (d) to eliminate any mistaken attribute detected in step (c) (Col. 6, lines 65-67 and Col. 7, lines 1-12, Rogers teaches filling holes in the mask by means of majority operation. The majority operation is a morphological operation in which each pixel in a binary image is turned on if a majority of its neighboring pixels are on. If the pixel is already on, it is left on.).
Therefore, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Dhungel in view of Mahmood in further view of Jurulu by filling holes in the mask that is taught by Rogers, to make the invention that generates an accurate silhouette (binary mask) of the breast which is used in later processing steps to decrease the time required for processing the mammogram image; thus, one of ordinary skilled in the art would be motivated to combine the references since if the holes in the mask are not filled in, the multiple erode steps may break the mask into disjointed sections (Rogers, Col. 7, lines 4-12).
Thus, the claimed subject matter would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention.
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 SYDNEY L BLACKSTEN whose telephone number is (571)272-7120. The examiner can normally be reached 8:30am-4:30pm.
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, Oneal Mistry can be reached at 313-446-4912. 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.
/SYDNEY L BLACKSTEN/Examiner, Art Unit 2674
/ONEAL R MISTRY/Supervisory Patent Examiner, Art Unit 2674
Prosecution Insights