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 1-20 are currently pending in the present application, with claims 1, 12, and 19 being independent.
Response to Amendments / Arguments
Applicant’s arguments, see Pg. 6-11, filed 07/23/2026, with respect to the rejection(s) of claim(s) 1-20 under 35 U.S.C. § 103 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 Therefore, the rejection has been withdrawn.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 in view of newly found prior art.
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 § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim(s) 1-11 is/are rejected under 35 USC 101 because the claimed invention is not directed to patentable subject matter. The claimed matter is directed to a judicial exception (i.e. abstract idea not integrated into a practical application) without significantly more.
Regarding claim 1, under Step 1 of the subject matter eligibility analysis, they are directed to a method claim (i.e., a process), which fall within the statutory categories of invention.
Under Step 2A Prong 1, the limitations as drafted of claim(s) 1, under its broadest reasonable interpretation, are directed to an abstract idea without significantly more. The instant invention is broadly directed to a method for training a model for unpaired image-to-image translation of medical images. Claim 1 recites the following (with emphasis added):
A method for training a model for unpaired Image-to-Image translation of medical images, the method comprising:
acquiring a plurality of real patient images;
acquiring a plurality of synthetic images of digital phantoms; and
training a model to transform the plurality of synthetic images of digital phantoms to resemble the plurality of real patient images, wherein the model is trained using at least one specialized loss function calculated between the transformed images and the plurality of synthetic images that ensures the transformed images retain original Hounsfield Unit (HU) values of the plurality of synthetic images.
It is further noted that the steps recited of calculating a loss function between respective image data constitutes a mathematical concept, can be considered a mathematical concept, directed to the mathematical relationships and comparisons between representations of the transformed and synthetic images for use in training the model. Thus, the limitations recite an abstract idea. (See MPEP 2106.04(a))
Under Step 2A Prong 2, claim 1 as a whole does not integrate the recited mathematical concept into a practical application.
Claim 1 limits the mathematical training operations to the technological environment of medical unpaired image-to-image translation. However, limiting the application of an abstract idea to a particular technological environment or field of use does not, without more, integrate the exception into a practical application. Although claim 1 requires that the specialized loss function ensures the transformed images retain original Hounsfield Unit (HU values), this limitation states the result to be achieved by the mathematical loss calculation. The claim does not recite a particular mathematical formulation or technological mechanism by which HU values are retained, nor does it recite how operation of the machine-learning architecture itself is technologically improved. Rather, the loss function is functionally defined according to the desired property of its resulting transformed images, and therefore, the claim does not recite a specific improvement in the operation of the machine-learning model, computer, or medical-imaging device itself (Recentive Analytics, Inc. v. Fox Corp., (Fed. Cir. 2025)).
Accordingly, the claim recites an abstract idea and these additional claim elements do not integrate the abstract idea into a practical application, because (1) they do not effect improvements to the functioning of a computer, or to any other technology or technical field (see MPEP 2106.05 (a)); (2) they do not apply or use the abstract idea to effect a particular treatment or prophylaxis for a disease or a medical condition (see the Vanda memo); (3) they do not apply the abstract idea with, or by use of, a particular machine (see MPEP 2106.05 (b)); (4) they do not effect a transformation or reduction of a particular article to a different state or thing (see MPEP 2106.05 (c)); (5) they do not apply or use the abstract idea in some other meaningful way beyond generally linking the use of the identified abstract idea to a particular technological environment, such that the claim as a whole is more than a drafting effort designated to monopolize the exception (see MPEP 2106.05 (e) and the Vanda memo). Therefore, per Step 2A, Prong Two, the claim is directed to an abstract idea not integrated into a practical application.
Under Step 2B, claim 1 does not include additional elements that are sufficient to amount to significantly more than the abstract idea judicial exception because when considered individually, these additional claim elements serve merely to implement the abstract idea using computer components performing computer functions. They do not constitute “Improvements to the Functioning of a Computer or to Any Other Technology or Technical Field” (MPEP 2106.05(a)).
The additional limitations of acquiring image data, employing a machine-learning model or generator network, and implementing the mathematical loss calculation through an image-to-image translation architecture are merely recited computational components used as tools for performing the mathematical operations. The use of existing machine-learning technology to perform a task with increased speed or efficiency does not itself establish an improvement in computer or machine-learning technology, merely applying machine-learning techniques in a particular field does not supply significantly more where the claim remains directed to performing the underlying abstract idea (Recentive Analytics, Inc. v. Fox Corp., (Fed. Cir. 2025)). Considered individually and as a whole, the additional elements do not amount to significantly more than the recited abstract idea
Further, Step 2B of the analysis takes into consideration all dependent claims as well, both individually and as a combination.
Claim(s) 2-3 merely adds a further limitation of data acquisition upon which the subsequent mathematical operation is performed, and does not recite an improvement in the operation of simulators and CT imaging systems itself. The additional element’s individual and combined significance is still not heavier than the abstract concepts at the core of the claimed invention, Therefore, also neither a practical application nor significantly more than the abstract idea.
Claim(s) 4-6 merely adds a further limitation to the model of claim 1 to a generator network trained using an adversarial machine-learning process, including CycleGAN or StarGAN architectures. These additional limitations further specify the machine-learning environment in which the mathematical processing is performed, but do not themselves recite a technological improvement to the operation of the computer or MLM. The additional element’s individual and combined significance is still not heavier than the abstract concepts at the core of the claimed invention, Therefore, also neither a practical application nor significantly more than the abstract idea.
Claim(s) 7-11 merely adds a further limitation to specify the particular loss calculation or comparison used during model training, further limiting the recited mathematical processing. The additional element’s individual and combined significance is still not heavier than the abstract concepts at the core of the claimed invention, Therefore, also neither a practical application nor significantly more than the abstract idea.
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) 1-6, and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liang et al. (US 20220084173), hereinafter referred to as “Liang”, in view of Abadi, Ehsan, et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in further view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”.
Regarding claim 1, Liang discloses a method for training a model for unpaired Image-to-Image translation of medical images (Par. 0011; Fixed-Point GAN…image-to-image translation…field of medical image processing. Par. 0039; a modified GAN… (1) handling unpaired images; (2) translating any image to a target domain reQiuring no source domain; (3) performing an identity transformation during same-domain translation; (4) performing minimal image transformation for cross-domain translation; and (5) scaling efficiently to multi-domain translation. See more in Par. 0065-0069), the method comprising:
acquiring a plurality of real patient images (Par. 0100-0101; BRATS 2013 dataset consists of…real brain MR images…),
acquiring a plurality of synthetic images (Par. 0100-0101; BRATS 2013 dataset consists of synthetic…brain MR images…)
and training a model to transform the plurality of synthetic images to resemble the plurality of real patient images (Par. 0057; Because the Fixed-Point GAN is trained using unpaired images, cycle-consistency is further utilized as set forth by Equation 5, so as to ensure that the generated images are close to the input images in both cross-domain (FIG. 3B, operation “E”) and same-domain (FIG. 3C operation “H”) translation learning. Par. 0104; …(1) disease-to-healthy and (2) healthy-to-healthy translations at the testing stage…Fixed-Point GAN's ability to perform both local and identical transformations),
Liang does not appear to explicitly disclose digital phantoms.
In the same art of medical imaging systems, Abadi discloses digital phantoms (Pg. 042805-2 - 042805-5, Section 2, Fig. 2, and Fig. 3; computational, anthropomorphic phantoms…procedurally generated phantoms…patient-based phantoms).
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 synthetic images generated from digital phantoms, as taught by Abadi, as input data for the unpaired image-to-image translation method disclosed by Liang. The motivation lies in the advantage of computational phantoms being a known, conventional way for providing a “gold standard” or “ground truth” from which to quantitatively evaluate and improve imaging devices and techniques (Abadi Section 2; The advantage of computational phantoms is that, unlike actual patients, their exact anatomy is known, providing a “gold standard” or “ground truth” from which to quantitatively evaluate and improve imaging devices and techniques… The user knows precisely what simulated images should reveal in terms of organ volumes or boundaries, tumor locations, sizes, shapes, extent and frequency of motion, presence and location of disease indicators, etc. The dose to the organs and structures from different procedures can also be calculated to assess patient risk from radiation exposure…For VCTs, it is essential to have computational phantoms that are realistic so that simulated results emulate what should occur in actual subjects. Phantoms must realistically model patient anatomy and physiology including the geometry of the organs and structures, the material properties of the tissues, patient motions, blood flow or contrast perfusion, alterations of the anatomy due to disease, and any other factors that could affect medical imaging…), and yielding predictable results in diversity and reliability of training data for medical imaging.
Liang in view of Abadi does not disclose wherein the model is trained using at least one specialized loss function calculated between the transformed images and the plurality of synthetic images that ensures the transformed images retain original Hounsfield Unit (HU) values of the plurality of synthetic images.
In the same art of unpaired deep-learning medical image-to-image translation, Rossi discloses wherein the model is trained using at least one specialized loss function calculated between the transformed images and the plurality of synthetic images that ensures the transformed images retain original Hounsfield Unit (HU) values of the plurality of synthetic images (Fig. 4 and Pg. 6-7, Section 2.4; the cGAN structure incorporated…GCT was used to generate Sct from CBCT, while GCBCT was used to generate synthetic CBCT (sCBCT) from CT…generator loss functions…Pg. 14, Appendix A; in the unsupervised training, the generator loss functions included…adversarial loss, cycle consistency loss, and identity loss…identity mapping loss, was introduced in order to preserve HU values between real CT and sCT and between real CBCT and sCBCT…Eq. (A5) and (A6))
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Rossi’s identity mapping loss into the unpaired medical image translation model of Liang as applied to Abadi’s phantom-generated synthetic images. As taught in Rossi, doing so allows preservation of HU values in unsupervised CT image translation, yielding predictable results in maintaining quantitative radiological information while transforming image appearance and improving reliability of generated medical images (Rossi Section 4.1; more reliable in preserving the anatomical structure and the up-to-date information contained in the CBCT scan…Appendix A; preserve HU values between real CT and sCT and between real CBCT and sCBCT).
Regarding claim 2, Liang in view of Abadi in further view of Rossi discloses the method of claim 1, and further discloses wherein the plurality of real patient images are provided by scanning a patient using a CT medical imaging system (Liang Par. 0100-0101; BRATS 2013 dataset consists of…real brain MR images. Par. 0103; 121 computed tomography pulmonary angiography (CTPA) scans. Par. 0165; diseased medical scan of a patient).
Liang, Abadi, and Rossi are combined for the reason set forth above with respect to claim 1.
Regarding claim 3, Liang in view of Abadi in further view of Rossi discloses the method of claim 1, and further discloses wherein the plurality of synthetic images are provided by scanning the digital phantoms (Liang Par. 0100-0101; BRATS 2013 dataset consists of synthetic…brain MR images…MR imaging sequence (FLAIR) for all patients in both HG and LG categories, resulting in a total of 9,050 synthetic MR slices…).
Liang does not disclose using a medical imaging system simulator.
In the same art of medical imaging systems, Abadi discloses using a medical imaging system simulator (Pg. 042805-2, Section 1; Virtual clinical trials (VCTs)…simulation experiments could be in the context of human models being imaged with imaging devices. Pg. 042805-3, Section 2.1; Phantoms are first constructed by defining objects to represent the necessary organs and structures of a given subject…for input into corresponding imaging simulation).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the medical imaging system simulator taught by Abadi into Liang’s Fixed-Point-GAN training. Incorporating a simulation system enables controlled generating/scanning of synthetic images from digital phantoms, allowing for tailored variations in imaging parameters. The motivation lies in the advantage of improving the quality and realism of training data for unpaired image-to-image translation in medical imaging (Section 2; For VCTs, it is essential to have computational phantoms that are realistic so that simulated results emulate what should occur in actual subjects. Phantoms must realistically model patient anatomy and physiology including the geometry of the organs and structures, the material properties of the tissues, patient motions, blood flow or contrast perfusion, alterations of the anatomy due to disease, and any other factors that could affect medical imaging).
Regarding claim 4, Liang in view of Abadi in further view of Rossi discloses the method of claim 1, and further discloses wherein the model comprises a generator network trained using an adversarial machine learning process (Liang Par. 0011-0012; Fixed-Point GAN…implementing fixed-point image-to-image translation using improved Generative Adversarial Networks (GANs))
Liang, Abadi, and Rossi are combined for the reason set forth above with respect to claim 1.
Regarding claim 5, Liang in view of Abadi in further view of Rossi discloses the method of claim 4 and further discloses wherein the generator network is trained using a CycleGAN architecture (Liang Par. 0132; use of CycleGAN provides for improvements in unpaired image-to-image translations via cycle consistency. Par. 0063; cycle consistency loss. Fig. 5A and Par. 0087; The experiment compared the Fixed-Point GAN with CycleGAN…CycleGAN can be utilized only to translate images containing two domains, comparing it as a baseline provides more insight into the performance of the Fixed-Point GAN as a scalable alternative).
Liang, Abadi, and Rossi are combined for the reason set forth above with respect to claim 1.
Regarding claim 6, Liang in view of Abadi in further view of Rossi discloses the method of claim 4, and further discloses wherein the generator network is trained using a STARGAN architecture (Liang Par. 0076; StarGAN was used as the baseline to provide multi-domain image-to-image translation. Fig. 4A-4B, 5A and Par. 0087; The experiment compared the Fixed-Point GAN with…StarGAN…allowed the study of the effect of fixed-point translation. Par. 0109; comparison with StarGAN reveals the effect of the proposed fixed-point translation learning. Par. 0123; the same generator and discriminator architectures as the public implementation of StarGAN were used. All models were trained using the Adam optimizer with a learning rate of 1e.sup.-4 for both the generator and discriminator across all experiments).
Liang, Abadi, and Rossi are combined for the reason set forth above with respect to claim 1.
Regarding claim 11, Liang in view of Abadi in further view of Rossi discloses the method of claim 1, and further discloses wherein the specialized loss function comprises a comparison between the output image and an annotated image (Liang Par. 0052; 1) distinguish between real images and fake (e.g., translated or manipulated) images…training begins with providing the discriminator network a batch of random real images from the dataset as input. Par. 0055; if the input image has eyeglasses, then c.sub.x=with eyeglasses and c.sub.y=without eyeglasses…generator is trained to generate images in the correct domain. Par. 0057; ensure that the generated images are close to the input images in both cross-domain (FIG. 3B, operation “E”) and same-domain (FIG. 3C operation “H”) translation learning). Examiner's note; domain-labeled images functionally act as ground-truth annotations (e.g., with glasses vs without glasses)
Liang, Abadi, and Rossi are combined for the reason set forth above with respect to claim 1.
Claim(s) 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liang et al. (US 20220084173), hereinafter referred to as “Liang”, in view of Abadi et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in further view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”, and in further view of Sakboonyarat et al. Discriminative Image Enhancement for Robust Cascaded Segmentation of CT Images. ECTI Transactions on Computer and Information Technology (ECTI-CIT) (2021), hereinafter referred to as “Sakboonyarat”.
Regarding claim 7, Liang in view of Abadi in further view of Rossi discloses the method of claim 1 but does not disclose wherein the specialized loss function comprises a loss value based on a comparison of HU value histograms.
In the same art of medical imaging systems, Sakboonyarat discloses wherein the specialized loss function comprises a loss value based on a comparison of HU value histograms (Pg. 153-155, Section 3.1; We create an HU histogram and assume that the HU with maximum frequency in the bounding box of the target region represents typical HU values of the target…Section 3.3; HU values from all voxels inside these 2D bounding boxes are accumulated to create the HU histogram…Fig. 5-6. Pg. 160, Section 4.4; directly related to the loss function utilized in model training).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to include loss values of HU value histograms, as taught by Sakboonyarat, in Liang, Abadi, and Rossi’s medical/CT imaging system. Hu value histogram is known tool used in CT scan imaging to help doctors quantify tissue characteristics and diagnose conditions, therefore, comparing input and generated images values yields the predictable result of ensuring the quality of the output image doesn’t deviate from the true norm in the diagnosis.
Claim(s) 8 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liang et al. (US 20220084173), hereinafter referred to as “Liang”, in view of Abadi et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in further view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”, and in further view of Tang et al. (US 20250166247), hereinafter referred to as “Tang”.
Regarding claim 8, Liang in view of Abadi in further view of Rossi discloses the method of claim 1 but does not disclose wherein the specialized loss function comprises calculating a region of interest loss.
In the same art of medical imaging, Tang discloses wherein the specialized loss function comprises calculating a region of interest loss (Par. 0054-0055; loss function of the vessel ROI…the loss function can be formulated based on the gradient to describe edge information within the vessel ROI).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the region of interest loss function as taught by Tang into the unpaired image-to-image translation model of Liang, Abadi, and Rossi’s combined system. The motivation lies in the advantage of localized fidelity in synthetic medical images, especially in diagnosing specific or significant areas.
Claim(s) 9 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liang et al. (US 20220084173), hereinafter referred to as “Liang”, in view of Abadi, Ehsan, et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in further view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”, and in further view of Zheng et al. (WO 2022120758), hereinafter referred to as “Zheng”.
Regarding claim 9, Liang in view of Abadi in further view of Rossi discloses the method of claim 1 but does not disclose wherein the specialized loss function comprises a feature matching loss.
In the same art of medical imaging, Zheng discloses wherein the specialized loss function comprises a feature matching loss (Pg. 8, Formula (4); the loss function of the conjugate generative adversarial network further includes a feature matching loss function).
It would have been obvious to a person of ordinary skill in art, before the effective filing date of the claimed invention, to incorporate the feature matching loss function as taught by Zheng into the unpaired image-to-image translation model of Liang, Abadi, and Rossi’s combined system. The motivation lies in the advantage of preserving semantic features and enhancing structural/diagnostic consistency in medical images.
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Liang et al. (US 20220084173), hereinafter referred to as “Liang”, in view of Abadi, Ehsan, et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in further view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”, and in further view of Guendel et al. (CN 113971657), hereinafter referred to as “Guendel”.
Regarding claim 10, Liang in view of Abadi in further view of Rossi discloses the method of claim 1, but does not disclose wherein the specialized loss function comprises a loss that enforces regularization or a physical simulation consistency.
In the same art of medical imaging, Guendel discloses wherein the specialized loss function comprises a loss that enforces regularization (Par. 0004; Machine training uses a loss function that includes regularization. Regularization is noise regularization and/or correlation regularization).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate the regularization loss function as taught by Guendel into the unpaired image-to-image translation model of Liang, Abadi, and Rossi’s combined system. The motivation lies in the advantage of enhancing image quality by reducing noise in order to obtain accurate and realistic medical images.
Claim(s) 12-17, 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Abadi et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”.
Regarding claim 12, Abadi discloses a system for performing a virtual clinical trial (VCTs in medical imaging), the system comprising:
a plurality of virtual digital phantoms and/or physical phantoms (Pg. 042805-2 - 042805-5, Section 2, Fig. 2, and Fig. 3; computational, anthropomorphic phantoms…procedurally generated phantoms…patient-based phantoms) wherein each of the virtual digital phantoms and/or physical phantoms includes one or more (Pg. 042805-2, Section 2; The advantage of computational phantoms is that, unlike actual patients, their exact anatomy is known, providing a "gold standard" or "ground truth" from which to quantitatively evaluate and improve imaging devices and techniques…Section 4; known ground truths),
a virtual imaging simulator configured to generate a virtual image from each of the plurality of virtual digital phantoms and/or physical phantoms (Pg. 042805-9 - 042805-15, Section 3; simulators of the imaging system to "virtually image" the virtual subjects…Section 3.1; to generate the simulated images, the acQiusition geometry, system components, and computational phantoms are input to an x-ray interaction simulation framework…Section 3.2 and Fig. 10; realistic simulated PET and SPECT images generated from different human phantoms…Section 3.3; Realistic 3-D MRI simulations of computational phantoms…Section 3.4 and Fig. 11; simulation creates a realistic scattering field from the complex numerical breast phantom).
Abadi does not disclose and a model configured for unpaired Image-to-Image translation, the model configured to transform the virtual images to resemble real patient images while maintaining the one or more ground truth HU values using at least one specialized loss function calculated between the transformed virtual images and the virtual images.
In the same art of medical imaging, Rossi discloses a model configured for unpaired Image-to-Image translation (Pg. 3-4, Section 1.1; The literature presents two main paradigms to train CNN for image-to-image translation: supervised and unsupervised training…), the model configured to transform the virtual images to resemble real patient images while maintaining the one or more ground truth HU values using at least one specialized loss function calculated between the transformed virtual images and the virtual images (Fig. 4 and Pg. 6-7, Section 2.4; the cGAN structure incorporated…GCT was used to generate Sct from CBCT, while GCBCT was used to generate synthetic CBCT (sCBCT) from CT…generator loss functions…Pg. 14, Appendix A; in the unsupervised training, the generator loss functions included…adversarial loss, cycle consistency loss, and identity loss…identity mapping loss, was introduced in order to preserve HU values between real CT and sCT and between real CBCT and sCBCT…Eq. (A5) and (A6)).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate HU-preserving image translation techniques taught by Rossi into the virtual clinical trials of Abadi. Integrating Rossi’s cGAN with identity mapping loss into a medical simulation system would enable the transformation of simulated images to resemble real patient images, allowing enhanced realism, improved reliability of generated medical images, and increased utility of VCTs. The combination provides predictable results in improving image fidelity and domain alignment of medical imaging (Rossi Section 3.2; resulting in more realistic and reliable contours with respect to the original CBCT…Section 4.1; more reliable in preserving the anatomical structure and the up-to-date information contained in the CBCT scan…Appendix A; preserve HU values between real CT and sCT and between real CBCT and sCBCT).
Regarding claim 13, Abadi in view of Rossi discloses the system of claim 12, and further discloses wherein the plurality of virtual digital phantoms are XCAT models (Abadi Fig. 2, 3, 6, and 8; XCAT phantoms).
Abadi and Rossi are combined for the reason set forth above with respect to claim 12.
Regarding claim 14, Abadi in view of Rossi discloses the system of claim 12, and further discloses wherein the virtual imaging simulator is configured to simulate a CT scan of the plurality of virtual digital phantoms and/or physical phantoms (Abadi Pg. 024805-2; VCT, the human subject is replaced with a virtual digital phantom, the imaging system with a virtual simulated scanner…imaging data of a computer phantom can be generated using a computerized scanner model…Section 3.1; Fig. 9 shows examples of scanner-specific simulated images of mammography, tomosynthesis, and CT…Section 3.2; SPECT scanner simulations…Section 5.2; CT imaging).
Abadi and Rossi are combined for the reason set forth above with respect to claim 12.
Regarding claim 15, Abadi in view of Rossi discloses the system of claim 12, but Abadi does not disclose wherein the model comprises a GAN based architecture.
Rossi discloses wherein the model comprises a GAN based architecture (Pg. 6-7, Section 2.4; cycle Generative Adversarial Network (cGAN)).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to adopt the GAN-based architecture of Rossi in Abadi’s virtual clinical trial system. Rossi’s GAN-based architecture is well-suited for transforming unpaired image domains and is widely used in medical imaging systems (Pg. 3, Section 1.1; unsupervised training enabled the possibility to use unpaired data for image-to-image translation [34]. The most common architecture for this kind of training is cycle Generative Adversarial Network (cGAN)).
Regarding claim 16, Abadi in view of Rossi discloses the system of claim 12, but Abadi does not disclose wherein the model comprises a Generative AI based architecture.
Rossi discloses wherein the model comprises a Generative AI based architecture (Pg. 6-7, Section 2.4; cycle Generative Adversarial Network (cGAN)).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to implement Rossi’s Generative AI techniques in Abadi’s virtual clinical trial system. Providing machine-learning-based generative method trained on unpaired datasets is ideal for enhancing synthetic image realism, yielding predictable results in allowing transformation of virtual phantom outputs into clinical imaging for post-process use.
Regarding claim 17, Abadi in view of Rossi discloses the system of claim 12, but Abadi does not disclose wherein the model is trained using an additional loss function that maintains the one or more ground truth HU values.
Rossi discloses wherein the model is trained using an additional loss function that maintains the one or more ground truth HU values (Fig. 4 and Pg. 6-7, Section 2.4; the cGAN structure incorporated…GCT was used to generate Sct from CBCT, while GCBCT was used to generate synthetic CBCT (sCBCT) from CT…generator loss functions…Pg. 14, Appendix A; in the unsupervised training, the generator loss functions included…adversarial loss, cycle consistency loss, and identity loss…identity mapping loss, was introduced in order to preserve HU values between real CT and sCT and between real CBCT and sCBCT…Eq. (A5) and (A6)).
Abadi and Rossi are combined for the reasons set forth above with respect to claim 12.
Regarding claim 19, Abadi discloses A method for generating a synthetic medical image, the method comprising:
selecting a digital phantom (Pg. 042805-2 - 042805-5, Section 2, Fig. 2, and Fig. 3; computational, anthropomorphic phantoms…procedurally generated phantoms…patient-based phantoms. Fig. 2, 3, 6, and 8; XCAT phantoms),
inputting the digital phantom into a medical imaging simulator configured to generate a simulated image of the digital phantom (Pg. 042805-9 - 042805-15, Section 3; simulators of the imaging system to "virtually image" the virtual subjects…Section 3.1; to generate the simulated images, the acQiusition geometry, system components, and computational phantoms are input to an x-ray interaction simulation framework…Section 3.2 and Fig. 10; realistic simulated PET and SPECT images generated from different human phantoms…Section 3.3; Realistic 3-D MRI simulations of computational phantoms…Section 3.4 and Fig. 11; simulation creates a realistic scattering field from the complex numerical breast phantom).
Abadi does not disclose inputting the simulated image into an unpaired image to image translation network configured to generate a realistic version of the simulated image, wherein the unpaired image to image translation network is trained using at least one specialized loss function calculated between the realistic version of the simulated image and the simulated image to ensure the realistic version of the simulated image retains original Hounsfield Unit (HU) values of the simulated image, and evaluating the realistic version of the simulated image.
In the same art of medical imaging, Rossi discloses inputting the simulated image into an unpaired image to image translation network configured to generate a realistic version of the simulated image (Pg. 3-4, Section 1.1; The literature presents two main paradigms to train CNN for image-to-image translation: supervised and unsupervised training. Supervised training requires paired images from two dominions (e.g., CBCT and CT) for model training. The input CBCT is processed by the model generating a synthetic CT(sCT), which is then compared to the corresponding ground-truth CT to minimize their pixel-by-pixel difference iteratively… unsupervised training enabled the possibility to use unpaired data for image-to-image translation…), wherein the unpaired image to image translation network is trained using at least one specialized loss function calculated between the realistic version of the simulated image and the simulated image to ensure the realistic version of the simulated image retains original Hounsfield Unit (HU) values of the simulated image (Fig. 4 and Pg. 6-7, Section 2.4; the cGAN structure incorporated…GCT was used to generate Sct from CBCT, while GCBCT was used to generate synthetic CBCT (sCBCT) from CT…generator loss functions…Pg. 14, Appendix A; in the unsupervised training, the generator loss functions included…adversarial loss, cycle consistency loss, and identity loss…identity mapping loss, was introduced in order to preserve HU values between real CT and sCT and between real CBCT and sCBCT…Eq. (A5) and (A6), and evaluating the realistic version of the simulated image (Fig. 6-7).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to input the simulated image as taught by Abadi into Rossi’s unpaired image-to-image translation network trained using a specialized loss function. The combination yields predictable results in enhancing image and diagnostic utility of simulated images, allowing synthetic output to resemble real-world scans in the field of medical imaging and AI-driven virtual clinical trials (Rossi Pg. 9, Section 3.2; resulting in more realistic and reliable contours with respect to the original CBCT).
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable Abadi et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”, and in further view of Sakboonyarat et al. Discriminative Image Enhancement for Robust Cascaded Segmentation of CT Images. ECTI Transactions on Computer and Information Technology (ECTI-CIT) (2021), hereinafter referred to as “Sakboonyarat”.
Regarding claim 18, Abadi in view of Rossi discloses the system of claim 17, but does not disclose wherein the additional loss function comprises a loss value based on a comparison of HU value histograms between original and generated images.
In the same art of medical imaging systems, Sakboonyarat discloses wherein the additional loss function comprises a loss value based on a comparison of HU value histograms between original and generated images (Pg. 154-155, Section 3.3; HU values from all voxels inside these 2D bounding boxes are accumulated to create the HU histogram…Fig. 5; Procedure for creating an HU histogram…Fig. 6; HU histograms of liver ground truth (red), accumulated 2D bounding boxes detected by RetinaNet (blue dotted line), and original CT slices containing the liver region (gray)…Pg. 160, Section 4.4; directly related to the loss function utilized in model training).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to include loss values of HU value histograms, as taught by Sakboonyarat, in Abadi and Rossi’s combined system. Hu value histogram is known tool used in CT scan imaging to help doctors quantify tissue characteristics and diagnose conditions, therefore, comparing input and generated images values yields the predictable result of ensuring the quality of the output image doesn’t deviate from the true norm in the diagnosis.
Claim(s) 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Abadi, Ehsan, et al. "Virtual clinical trials in medical imaging: a review." Journal of Medical Imaging 7.4 (2020): 042805-042805, hereinafter referred to as “Abadi”, in view of Rossi et al. Comparison of Supervised and Unsupervised Approaches for the Generation of Synthetic CT from Cone-Beam CT. Diagnostics 2021; 11: 1435. 2021, hereinafter referred to as “Rossi”, in further view of Wang et al. "TWIN-GPT: digital twins for clinical trials via large language model." ACM Transactions on Multimedia Computing, Communications and Applications (2024), hereinafter referred to as “Wang”.
Regarding claim 20, Abadi in view of Rossi discloses the method of claim 19, and further discloses the digital phantom (Liang Pg. 042805-2 - 042805-5, Section 2, Fig. 2, and Fig. 3; computational, anthropomorphic phantoms…procedurally generated phantoms…patient-based phantoms), medical imaging simulator (Liang Pg. 042805-9 - 042805-15, Section 3; simulators of the imaging system), and scan parameters (Liang Pg. 042805-2, Section 2; Imaging data of a computer phantom can be generated using a computerized scanner model under various scanning parameters or protocols, and the effects quantified in comparison with the known phantom).
Abadi in view of Liang does not disclose are selected or provided by a chatbot.
In the same art of AI models and clinical trials, Wang discloses selected or provided by a chatbot (Pg. 2, Section 1; TWIN-GPT is fine-tuned on a pre-trained LLM (ChatGPT [52]) on clinical trial datasets, so as to generate personalized digital twins for different patients).
It would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Wang’s chatbot into the medical imaging system of Abadi and Rossi. The motivation lies in the advantage of automating the medical imaging system by allowing a chatbot to assist in selecting or providing simulation inputs in order to streamline simulation configuration, personalize virtual patients, and increase scalability in clinical research.
Conclusion
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 /YANNA WU/ Primary Examiner, Art Unit 2615
Prosecution Insights