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 .
Application Status
This office action is responsive to the remarks filed 04/02/2026.
This action has been made NON-FINAL.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 04/02/2026 has been entered.
Response to Arguments
Applicant's arguments filed 04/02/2026 have been fully considered but they are not persuasive.
The Applicant alleges the following: “Guibas does not disclose a user-specified geometric curve representing geometric characteristics of anatomical structures, nor does Guibas disclose a neural network conditioned on such a curve as required by the claim language.” The examiner is not persuaded. Guibas teachings of “the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” in 15 Future Work section is the same as the Applicant’s claim language of “an access component that accesses a user-specified geometric curve.” Guibas goes on to disclose “representing geometric characteristics of one or more anatomical structures” in Figure 1; 3 Data section1. The examiner interprets Guibas teachings of “retina fundus image” in Figure 1 as being the same as the Applicant’s teachings of “anatomical structures.” Accordingly, the examiner maintains the rejection.
The Applicant alleges the following: “Guibas itself confirms that geometric curve representations are not part of the disclosed system. Specifically, Guibas describes conditioning on higher-level abstractions such as Bezier curves only as "future work." Accordingly, the reference does not disclose a system in which a neural network is conditioned on a user-specified geometric curve as presently claimed.” The examiner is not persuaded. Even though, a small portion of Guibas merely mentions “Future Work” on pages 7-8, the Guibas reference, still qualifies as prior art because it was publicly available before the effective filing date. Moreover, the claims, as written are broad enough to read on the disclosed language. Accordingly, the examiner maintains the rejection.
The Applicant alleges the following: “Dependent claim 2 further recites that the first deep learning neural network receives as input the user-specified geometric curve concatenated with a randomly generated array, and expressly requires that the user-specified geometric curve is a geometric representation independent of a pixel-based segmentation mask. Guibas does not disclose a geometric curve input that is independent of a pixel-based segmentation mask, nor does Guibas disclose concatenation of such a user-specified geometric curve with a randomly-generated array as claimed.” The examiner is not persuaded. Guibas teachings of “the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” in 15 Future Work section discloses the Applicant’s claim language of “wherein the user-specified geometric curve is a geometric representation independent.” Guibas goes on to disclose “a pixel-based segmentation mask” in 8 U-net section2. Moreover, Guibas teachings “Receptive fields after convolution are also concatenated with receptive fields in the decoding process” in 8 U-net section discloses “concatenated with a randomly-generated array.” Accordingly, the examiner maintains the rejection.
The Applicant alleges the following: “Claims 3 and 4 depend from claims 2 and 1, respectively, and likewise require conditioning based on the recited user-specified geometric curve. Since Guibas fails to disclose the limitations of the parent claims, anticipation of claims 3 and 4 necessarily fails.” The examiner is not persuaded. As mentioned above, Guibas teachings of “the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” in 15 Future Work section discloses the Applicant’s claim language of “wherein the user-specified geometric curve is a geometric representation independent.” Accordingly, the examiner maintains the rejection.
The Applicant alleges the following: “With respect to claim 9, like claim 2, Guibas does not disclose a geometric curve input that is independent of a pixel-based segmentation mask, nor does Guibas disclose concatenation of such a user-specified geometric curve with a randomly-generated array as claimed.” The examiner is not persuaded. As mentioned above, Guibas teachings of “the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” in 15 Future Work section discloses the Applicant’s claim language of “wherein the user-specified geometric curve is a geometric representation independent.” Guibas goes on to disclose “a pixel-based segmentation mask” in 8 U-net section3. Moreover, Guibas teachings “Receptive fields after convolution are also concatenated with receptive fields in the decoding process” in 8 U-net section discloses “concatenated with a randomly-generated array.” Accordingly, the examiner maintains the rejection.
The Applicant alleges the following: “Guibas does not disclose training datasets defined by geometric curves corresponding to medical images as expressly recited.” The examiner is not persuaded. Guibas teachings of “the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” in 15 Future Work section is the same as the Applicant’s claim language of “an access component that accesses a user-specified geometric curve.” Guibas teachings in Figure 1; 3 Data section4 discloses the Applicant’s claim language of “defined by geometric curves corresponding to medical images.” Moreover, Guibas teachings of “example generated segmentation masks and their closest mask in the training dataset” in Figure 4 discloses the Applicant’s claim language. MPEP § 2106 states Office personnel are to give claims their broadest reasonable interpretation in light of the supporting disclosure. In re Morris, 127 F.3d 1048, 1054-55, 44 USPQ2d 1023, 1027-28 (Fed Cir. 1997). Accordingly, the examiner maintains the rejection.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-20 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Guibas et al., “Synthetic Medical Images from Dual Generative Adversarial Networks”, arXiv:1709.01872v3, Jan. 8, 2018, pp. 1-9 (Year: 2018).
Claim 1:
Guibas discloses a system (See Abstract), comprising:
a processor that executes computer-executable components stored in a non-transitory computer-readable memory, the computer-executable components (“Computer-aided medical diagnosis” See Introduction section) comprising:
an access component that accesses a user-specified geometric curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) representing geometric characteristics of one or more anatomical structures (See Figure 1; 3 Data section5);
an inference component that generates, via execution of a first deep learning (See Introduction section) neural network conditioned (See 5 Generative Adversarial Network section) on the user-specified geometric curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section), a synthetic medical image (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset) whose visual characteristics are based on the user-specified geometric curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) wherein the synthetic medical image (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset) depicts one or more anatomical structures having lengths, shapes, positions, or orientations (See 12 Pipeline Validation section & 12 Discussion section) that correspond to the user-specified geometric curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section);
and a display component that renders the synthetic medical image on an electronic display (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset).
Claim 2:
Guibas discloses wherein the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network
section) receives as input the user-specified geometric curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) concatenated with a randomly-generated array (“Receptive fields after convolution are also concatenated with receptive fields in the decoding process” See 8 U-net section), wherein the user-specified geometric curve is a geometric representation independent of (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) a pixel-based segmentation mask (See 8 U-net section6) and wherein the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network
section) produces as output the synthetic medical image (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset).
Claim 3:
Guibas discloses wherein the user-specified geometric curve is further concatenated with a text conditioning or a sketch conditioning (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section).
Claim 4:
Guibas discloses wherein the user-specified geometric curve is a two-dimensional curve or a three-dimensional curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section).
Claim 5:
Guibas discloses wherein the access component accesses a training dataset (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10].
DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order
to find a mapping between the two classes of images.” See 7 Stage-II GAN section), and wherein the computer-executable components further comprise: a training component that trains the first (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10]. DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order to find a mapping between the two classes of images.” See 7 Stage-II GAN section) deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network
section) based on the training dataset (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10]. DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order to find a mapping between the two classes of images.” See 7 Stage-II GAN section).
Claim 6:
Guibas discloses wherein the training dataset (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10]. DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order to find a mapping between the two classes of images.” See 7 Stage-II GAN section) comprises a set of training medical images (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10]. DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order to find a mapping between the two classes of images.” See 7 Stage-II GAN section) and a set of training geometric curves (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) respectively corresponding to the set of training medical images (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10].
DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order
to find a mapping between the two classes of images.” See 7 Stage-II GAN section), and wherein the training component trains the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section) by:
selecting, from the training dataset, a training medical image and a training geometric curve that corresponds to the training medical image (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10].
DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order
to find a mapping between the two classes of images.” See 7 Stage-II GAN section);
iteratively applying noise to the training medical image, thereby yielding a first output (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10]. DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order to find a mapping between the two classes of images.” See 7 Stage-II GAN section);
executing the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section) on the first output and on the training geometric curve, thereby yielding a second output (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section);
and updating internal parameters of the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section), via backpropagation based on an error between the second output and the training medical image (“The noise vector is up sampled and the weights of G are learned through backpropagation, eventually producing data that is classified as real by the discriminator. Further, a key feature of GANs is the ability to produce a larger amount of images than the original dataset” See 5 Generative Adversarial Network section).
Claim 7:
Guibas discloses wherein the training dataset comprises a set of training medical images (“We trained the GAN in Stage-I with retinal vessel segmentations from the DRIVE database [10]. DRIVE contains forty pairs of retinal fundi images and vessel segmentation masks manually labeled by two experts” See 3 Data section AND “The Stage-II GAN is trained with corresponding pairs of real fundi and segmentations masks in order to find a mapping between the two classes of images.” See 7 Stage-II GAN section) and a set of training geometric curves (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) respectively corresponding to the set of training medical images (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset), and wherein the training component trains the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section) by:
selecting, from the training dataset, a training medical image and a training geometric curve (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section) that corresponds to the training medical image (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset);
executing a second deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section) on the training medical image (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset), thereby yielding a first output (See Figure 4: Example generated segmentation masks and their closest mask in the training dataset);
iteratively applying noise to the first output, thereby yielding a second output (“The noise vector is up sampled and the weights of G are learned through backpropagation, eventually producing data that is classified as real by the discriminator. Further, a key feature of GANs is the ability to produce a larger amount of images than the original dataset” See 5 Generative Adversarial Network section);
executing the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section) on the second output and on the training geometric curve, thereby yielding a third output (“the GAN in Stage-I producing an image, generative models trained with other representations such as bezier curves, 2D point clouds, or skeletons may be used to reduce dimensionality in the generation process” See 15 Future Work section);
and updating internal parameters of the first deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section) and of the second deep learning (See Introduction section) neural network (See 5 Generative Adversarial Network section), via backpropagation based on an error between the third output and the training medical image (“The noise vector is up sampled and the weights of G are learned through backpropagation, eventually producing data that is classified as real by the discriminator. Further, a key feature of GANs is the ability to produce a larger amount of images than the original dataset” See 5 Generative Adversarial Network section).
Claims 8-14:
Claims 8-14 are rejected on the same basis as claims 1-7.
Claims 15-19:
Claims 15-19 are rejected on the same basis as claims 1-3 and 5.
Claim 20:
Guibas discloses wherein the deep learning neural network comprises a stable diffusion denoiser or a stable diffusion decoder (“Receptive fields after convolution are also concatenated with receptive fields in the decoding process” See 8 U-net section).
Pertinent Art
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. US 20240135683 A1; US 20240233337 A9; US 12333786 B2 generate synthetic image(s) that are close to real image(s).
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SHEREE N BROWN whose telephone number is (571)272-4229. The examiner can normally be reached M-F 5:30-2:00 PM EST.
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, SAID BROOME can be reached at (571) 272-2931. 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.
/SHEREE N BROWN/Primary Examiner, Art Unit 2612 April 23, 2026
1 Figure 1 discloses “retina fundus image.”
2 Guibas recites “To evaluate the reliability of our synthetic data, we used it to train a u-net segmentation network which creates a segmentation mask given a photorealistic medical image.”
3 Guibas recites “To evaluate the reliability of our synthetic data, we used it to train a u-net segmentation network which creates a segmentation mask given a photorealistic medical image.”
4 Figure 1 discloses “retina fundus image.”
5 Figure 1 discloses “retina fundus image.”
6 Guibas recites “To evaluate the reliability of our synthetic data, we used it to train a u-net segmentation network which creates a segmentation mask given a photorealistic medical image.”
Prosecution Insights