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 .
Status of the Claims
Claims 1-16 are currently pending in the present application, with claims 1 and 9 being independent.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 06/04/2026 have been considered by the examiner.
Response to Amendments / Arguments
Applicant’s arguments with respect to claim(s) 1-16 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Regarding the remaining arguments: Applicant argues with respect to the amended claim language, which is fully addressed in the prior art rejections set forth below.
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.
Claim(s) 1-2, 5-10, and 13-16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Niu et al. "Noise entangled GAN for low-dose CT simulation." arXiv preprint arXiv:2102.09615 (2021)., hereinafter referred to as “Niu”, in view of Chen et al. "Image blind denoising with generative adversarial network-based noise modeling." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3155-3164. 2018, hereinafter referred to as “Chen”.
Regarding claim 1, Niu discloses an apparatus for denoising a medical image (NE-GAN, Section 2.3 Implementation details), comprising:
an image processing module configured to extract a noise component from a medical image for processing by inputting the medical image for the processing to a noise extraction deep learning model (Fig. 1-2 and Section 2.1 Generation of high-dose noise image; For the denoising scheme, the HDCT image is first forwarded into a denoising model to obtain a clean CT image, which is then subtracted from the input HDCT image to generate the high-dose noise image. The denoising model is trained by directly mapping high-dose CT image to the low-dose CT image, as shown in Fig. 2) trained in advance (Section 2.2, Right Column; After training, only the generator G is retained to simulate different levels of LDCT images given the clean image, the high-dose noise image, and the specific noise factor…Examiner’s note: after training to simulate different images clarifies that the model is trained in advanced), and generate a noise-removed image by subtracting the noise component from the medical image for the processing (Fig,. 1-2 and Section 2.1 Generation of high-dose noise image; the HDCT image is first forwarded into a denoising model to obtain a clean CT image, which is then subtracted from the input HDCT image to generate the high-dose noise image).
wherein the noise extraction deep learning model is trained using a simulation noise component image generated by a noise simulator and a simulation low-quality image generated based on the simulation noise component image as a pair (Fig. 1-2 and Section 2.1 Generation of high-dose noise image; the trained model can generate the denoised image…The denoising scheme can extract the real prior information from the real HDCT images, which are then transformed to LDCT images with specific noise level by NE-GAN…NE-GAN takes the simulated noise and the HDCT image as inputs to generate a set of LDCT images with different levels of noise…).
using training data including a set of normal-quality medical images for training (Section 2.1; high-dose noise image…simulated noise and the HDCT image as inputs…) and either a noise component image for training or a low-quality medical image for training (Fig. 1-2 and Section 2.1 Generation of high-dose noise image; the trained model can generate the denoised image…NE-GAN takes the simulated noise and the HDCT image as inputs to generate a set of LDCT images with different levels of noise…),
(Section 2.1; high-dose noise image…simulated noise and the HDCT image as inputs…).
Niu does not disclose wherein the noise simulator is trained in advance
wherein the trained noise simulator is configured to generate the simulation noise component image from
wherein the noise extraction deep learning model is trained using a pair of the simulation noise component image generated by the trained noise simulator and the simulation low-quality image generated based on the simulation noise component image.
In the same art of deep-learning image denoising models for supervised learning, Chen discloses wherein the noise simulator is trained in advance (Section 3.1.2; In the proposed method, a GAN is adopted to estimate the noise distribution over a set of approximate noise blocks V…The trained GAN model is used to generate noise samples for augmenting V and finally a larger set V' = {v1', v2', …, vw'} is obtained. Section 3.1.1; a set of approximate noise blocks (or patches), say V, need to be extracted first from the parts with weak background in the given noisy images…the set of approximate noise blocks V = {v1, v2, …, vt} can be derived by vi = si- Mean(si)…Section 3.1.2; model the noise distribution over these extracted blocks V, and then to generate more noise data…for the training of CNN…),
wherein the trained noise simulator is configured to generate the simulation noise component image (Fig. 3; x~ is a noise block generated by the generative network, and x is a noise block extracted from noisy images. Section 3.1.2; The trained GAN model is used to generate noise samples…), and
wherein the noise extraction deep learning model is trained using a pair of the simulation noise component image generated by the trained noise simulator (Section 3.2; The CNN is regarded as a single residual unit to predict the residual image, i.e., the difference between the input noisy image and the latent clean image. The objective function to be minimized is defined as Eq. (4), where Θ is the parameters of the network, N is the size of training data, yi is a noisy image, xi is the ground truth. Fig. 4; The input is a noisy image yi, and the output is the different R(yi; Θ) between the input and the latent clean image… Examiner's note: yi is the generated noisy/low-quality image input. yi-xi is the noise residual target.) and the simulation low-quality image generated based on the simulation noise component image (Fig. 2 and Section 3.2; Noise blocks V' are randomly added to patches in X to obtain Y = {y1, y2, …, yf}, where y1 = xj +vk'…).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to use the known GAN-based noise model training technique of Chen using Niu's CT images and corresponding noise information. Doing so allows the modelling of realistic CT noise for subsequent supervised denoising. Chen teaches generating simulated noise with a trained generative model, constructing corresponding simulated noisy images to form paired training data, and training a denoising network using the paired data. Therefore, applying Chen's known supervised training approach to Niu's simulated CT noise generation would merely be a use of a known training technique to improve the accuracy and effectiveness of the learned noise extraction model, further improving denoising performance and generalization.
Regarding claim 2, Niu discloses the apparatus of claim 1, and further discloses wherein the simulation low-quality image is generated by combining the simulation noise component image and a normal-quality medical image (Fig. 3 and Section 2.2 Noise Entangled GAN, Right Column; NE-GAN consists of generator G and a set of discriminators…the generator G is a encoder-decoder network that takes a clean CT image and a noise image scaled with a noise factor as inputs, and outputs a LDCT image corresponding to the input noise factor…)
Regarding claim 5, Niu discloses the apparatus of claim 1, and further discloses wherein the noise simulator is configured to generate the simulation noise component image by inputting the set of normal-quality medical images to a generative adversarial model trained in advance (Section 2 Methods; we propose to simulate an LDCT image through two steps: the first step is to generate a clean image…and the second step is to generate different levels of LDCT images by entangling the high-dose noise component scaled with a specific noise factor into the clean CT image. Section 2.2 Noise Entangled GAN, Right Column; NE-GAN consists of generator G and a set of discriminators…the generator G is a encoder-decoder network that takes a clean CT image and a noise image scaled with a noise factor as inputs, and outputs a LDCT image corresponding to the input noise factor)
Regarding claim 6, Niu discloses the apparatus of claim 5, and further discloses wherein the generative adversarial model is repeatedly trained (Section 3.1 Dataset; we used a multi-dose of real CT image dataset from [6], in which the CT images were collected from anonymous cadavers and each of them was repeatedly scanned…sub-dataset that contains 261 groups of CT images for training and 251 groups of CT images for testing…Section 3.2 Results on simulated dataset; we repeatedly generated HDCT noise image with CatSim and simulated the LDCT images with NE-GAN by 50 times) to minimize a loss due to difference between a noise component image for training and a simulation noise component image generated by inputting a set of normal-quality medical images for training to the generative adversarial model (Section 2.2 Noise Entangled GAN; The loss function is: …first two items in the loss function are the adversarial losses…and train G to minimize the probability of assigning the correct label for D, the third item is a data fidelity loss…and the fourth item is a reconstruction loss…).
Regarding claim 7, Niu discloses the apparatus of claim 1, and further discloses wherein the noise simulator is configured to generate a set of low-quality medical images by inputting a set of normal-quality medical images to a generative adversarial model trained in advance (Section 2 Methods; we propose to simulate an LDCT image through two steps: the first step is to generate a clean image…and the second step is to generate different levels of LDCT images by entangling the high-dose noise component scaled with a specific noise factor into the clean CT image. Section 2.2 Noise Entangled GAN, Right Column; NE-GAN consists of generator G and a set of discriminators…the generator G is a encoder-decoder network that takes a clean CT image and a noise image scaled with a noise factor as inputs, and outputs a LDCT image corresponding to the input noise factor…To train NE-GAN, we need a set of training samples…where xi^0 and ni^0 denote the clean CT image and high-dose image respectively…), and generate the simulation noise component image by subtracting the set of low-quality medical images from the set of normal-quality medical images (Section 2.1 Generation of high-dose noise image; For the denoising scheme, the HDCT image is first forwarded into a denoising model to obtain a clean CT image, which is then subtracted from the input HDCT image to generate the high-dose noise image. The denoising model is trained by directly mapping high-dose CT image to the low-dose CT image, as shown in Fig. 2).
Regarding claim 8, Niu discloses the apparatus of claim 7, and further discloses wherein the generative adversarial model is repeatedly trained to minimize a loss due to difference between a low-quality medical image for training and a simulation low-quality medical image generated by inputting a set of normal-quality medical images for training to the generative adversarial model (Section 2.2 Noise Entangled GAN; To train NE-GAN, we need a set of training samples… where xi^0 and ni^0 denotes the clean CT image and high-dose image respectively…The loss function is:…the first two items in the loss function are the adversarial losses that train D to maximize the probability of assigning the correct label to both real LDCT images and the generated ones from G…).
Regarding claim 9, claim 9 is the method claim of apparatus claim 1, and is accordingly rejected using substantially similar rationale as to that which is set for with respect to claim 1.
Regarding claim 10, claim 10 has similar limitations as of claim 2, except it is a method claim, therefore it is rejected under the same rationale as claim 2.
Regarding claim 13, claim 13 has similar limitations as of claim 5, except it is a method claim, therefore it is rejected under the same rationale as claim 5.
Regarding claim 14, claim 14 has similar limitations as of claim 6, except it is a method claim, therefore it is rejected under the same rationale as claim 6.
Regarding claim 15, claim 15 has similar limitations as of claim 7, except it is a method claim, therefore it is rejected under the same rationale as claim 7.
Regarding claim 16, claim 16 has similar limitations as of claim 8, except it is a method claim, therefore it is rejected under the same rationale as claim 8.
Claim(s) 3-4 and 11-12 is/are rejected under 35 U.S.C. 103 as being unpatentable over Niu et al. "Noise entangled GAN for low-dose CT simulation." arXiv preprint arXiv:2102.09615 (2021)., hereinafter referred to as “Niu”, in view of Chen et al. "Image blind denoising with generative adversarial network-based noise modeling." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3155-3164. 2018, hereinafter referred to as “Chen”, and in further view of Yang et al. "Low-dose CT denoising via sinogram inner-structure transformer." IEEE transactions on medical imaging 42, no. 4 (2022): 910-921, hereinafter referred to as “Yang”.
Regarding claim 3, Niu in view of Chen discloses the apparatus of claim 1, and Niu further discloses wherein the noise simulator is configured to generate the simulation noise component image by inputting the set of normal-quality medical images (Section 2 Methods; to generate a clean image. Section 2.2 Noise Entangled GAN, Right Column; NE-GAN consists of generator G and a set of discriminators…the generator G is a encoder-decoder network that takes a clean CT image and a noise image scaled with a noise factor as inputs), (Section 2.2 Noise Entangled GAN; To train NE-GAN, we need a set of training samples…S is number of noise levels or discriminators, kj denotes the noise factor that is a positive real number and a larger value corresponds to a higher noise
level or lower image quality).
Niu in view of Chen does not disclose of which a domain is converted into a sinogram domain
In the same art of LDCT techniques, Yang discloses (Fig. 2; sinogram noise and image noise and Section III.B; we propose the sinogram transformer module for sinogram domain denoising),
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to apply the sinogram-domain processing techniques as taught by Yang to the CT image processing framework of Niu and Chen’s combined system. Doing so preserves structural information related to the CT acquisition process and improve noise modeling and denoising performance (Yang Section I; Considering that noise appears randomly (e.g, Poisson Distribution) in the whole imaging procedure, the conjugate pairs obtaining different noise would evidently break this structure. By maintaining this sinogram inner-structure, we can effectively restrain the noise and improve the image quality. Therefore, exploring the inner-structure of sinogram is of great importance for sinogram domain denoising… Designing a loss function based on sinogram inner-structure for network training in the sinogram domain can effectively maintain this structure and restrain noise). Incorporating such projection-domain representations into noise simulation techniques would’ve yielded predictable results in improved accuracy and realism of the generated CT images.
Regarding claim 4, Niu in view of Chen and in further view of Yang discloses the apparatus of claim 3, and Niu further discloses wherein the noise generation model is provided as a model, of which parameters are varied depending on training, and repeatedly trained (Section 3.1 Dataset; we used a multi-dose of real CT image dataset from [6], in which the CT images were collected from anonymous cadavers and each of them was repeatedly scanned…sub-dataset that contains 261 groups of CT images for training and 251 groups of CT images for testing…Section 3.2 Results on simulated dataset; we repeatedly generated HDCT noise image with CatSim and simulated the LDCT images with NE-GAN by 50 times) to minimize a loss due to difference between a noise component image for training and the generated simulation noise component image (Section 2.2 Noise Entangled GAN; The loss function is: …first two items in the loss function are the adversarial losses…and train G to minimize the probability of assigning the correct label for D, the third item is a data fidelity loss…and the fourth item is a reconstruction loss…After training, only the generator G is retained to simulate different levels of LDCT images given the clean image, the high-dose noise image, and the specific noise factor…).
Niu, Chen, and Yang are combined for the reason set forth above with respect to claim 3.
Regarding claim 11, claim 11 has similar limitations as of claim 3, except it is a method claim, therefore it is rejected under the same rationale as claim 3.
Regarding claim 12, claim 12 has similar limitations as of claim 4, except it is a method claim, therefore it is rejected under the same rationale as claim 4.
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 JENNY NGAN TRAN whose telephone number is (571) 272-6888. The examiner can normally be reached Mon-Thurs 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner' s supervisor, Alicia Harrington can be reached at (571) 272-2330. 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.
/JENNY N TRAN/Examiner, Art Unit 2615 /ALICIA M HARRINGTON/Supervisory Patent Examiner, Art Unit 2615
Prosecution Insights