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 .
Priority
Application 17/962,734 filed on 10/10/2022, claims domestic priority data benefit of 63/254,812 filed on 10/12/2021.
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 03/23/2026 has been entered.
Response to Amendment
This office action is in response to amendments submitted on 03/23/2026 wherein claims 1, 3-7, 9-14, 16, 18, and 21 are pending and ready for examination. Claims 2, 17, and 20 have been canceled and claims 8, 15, and 19 have been previously canceled.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 9 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 9 which depends from claim 1: Applicant claims the limitation “the learned model is generated by machine learning using the first data as the third data.” where third data is “output learning data” which is “acquired based on first data” and where “first data” is “projection data” (claim 1) where the “projection data” is generated by the X-ray detector and signal processing circuitry (Written Specification, ¶ 0041). Examiner is confused as to “first data” which is “projection data,” can be “output learning data.” Claim 9 will be examined based on the merits as best understood.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 4-6, 9-14, and 16 and 20-21 are rejected under 35 U.S.C. 103 as being unpatentable over Zhao et al., hereinafter Zhao, “Sparse-View CT Reconstruction via Generative Adversarial Networks” downloaded from DOI: 10.1109/NSSMIC.2018.8824362 in view of Lee et al., hereinafter Lee, U.S. Pub. No. 2019/0139276 A1, as evidenced by Jacobs et al., “Deep Learning for Lung Cancer Detection on Screening CT Scans: Results of a Large-Scale Public Competition and an Observer Study with 11 Radiologists” downloaded from doi: 10.1148/ryai.2021210027.
Regarding independent claim 1 Zhao teaches:
“processing circuitry configured to (a) improve a quality of fourth data corresponding to a fourth number of views that is smaller than a first number of views” (Zhao, fig. 2, fig 3, § II. Method 1st – 2nd page, § III Dataset and Experiments, 3rd page: While Zhao does not explicitly teach processing circuitry, Zhao teaches using CNN and GAN which are machine learning techniques in order to manipulate data. A person of ordinary skill in the art would understand machine learning techniques such as CNN and GAN require at least a generic computer which would have processing circuitry which would include the interpolation network (fig. 2) generator and discriminator network (fig. 3). Additionally, Zhao teaches “We use sample images in the dataset as the original image . . .” (§ III, A. Dataset) where “sample images” “as the original image” discloses the “fourth data corresponding to a fourth number of views” and “the dataset” discloses the “first number of views.” The fourth number of views would be smaller than the first number of views as the “sample images” “as the original image” would be a smaller set that the “dataset” and thereby have a smaller number of views.
“by inputting the fourth data to a learned model generated by performing machine learning with second data corresponding to a second number of views as input learning data” (Zhao, § III Dataset and Experiments, 3rd page: Zhao teaches “Train a convolution neural network in the projection domain to complete the data” (§ I. Introduction) and “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the “sparse-view” discloses the “second data corresponding to a second number of views” and where the “sparse-view sinogram and the original full-view sinogram” being used as “the data pair to feed CNN” disclosing “second data” is used as “input learning data.”)
“and third data corresponding to a third number of views that is larger than the second number of views as output learning data” (Zhao, fig 1, § I. Introduction, 1st page: Zhao teaches “Reconstruct the images from the completed projection data using the FBP or WHLS methods” (§ I.) where the reconstructed images are the images after the FPB/WLS step and before the Estimation Networks (AN) step of fig 1. The reconstructed image discloses “third data corresponding to a third number of views.” Moreover, the WHLS (examiner believes this is a typographical error and should read WLS as supported by fig. 1) is part of a machine learning method of which the “third data corresponding to a third number of views” is output. The reconstructed image discloses “third data corresponding to a third number of views” which is “larger than the second number of views” as the “third view” is based on the “full view sinogram” and the “second number of views” is the “sparse-view sinogram” (see above).)
“the second data and the third data being acquired based on first data that is projection data of the first number of views” (Zhao, fig. 1, § III. Dataset and Experiments 3rd page: Zhao teaches “The data set used in the experiment is downloaded from the Data Science Bowl 2017. We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the CNN outputs the “full-view sinogram” where the “full-view sinogram” is reconstructed to produce a reconstructed image (third data) that is based on the “sparse-view sinogram” (second data) which is based on the “sample images” “as the original image” (fourth data) which comes from a “dataset” therefore is based on “first data.” Zhao teaches the “dataset” is a set of images from a computed tomography, CT, scan (“Data Science Bowl 2017”) as evidenced by Jacobs (see attached “Deep Learning for Lung Cancer Detection on Screening CT Scans” first page) disclosing the “first data” is “projection data.”)
“the second data is data corresponding to a sparser sparse view than the third data and is generated by downsampling the first data” (Zhao, § III. Dataset and Experiments 3rd page: Zhao teaches using a CNN which inputs a “sparse-view sinogram” (second data) and outputs a “full-view sinogram” which is reconstructed to generate a reconstructed image (third data) disclosing the “second data is data corresponding to a sparser sparse view than the third data.” Additionally, “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view” (§ III, A. Dataset). A person of ordinary skill in the art would understand Matlab, a programming and numeric computing platform that is commonly available, uses “downsampling” which can be used to create a sparse view image as evidenced by Taylor (see attached Matlab Help Center).)
Zhao does not teach:
“An X-ray diagnostic apparatus”
“display, on a display, the data whose quality has been improved by the learned model”
“the fourth data is data acquired by performing tomosynthesis imaging of comprising generating X-rays from an X-ray tube and detecting the X-rays that have passed through the subject using an X-ray detector”
Lee teaches:
“An X-ray diagnostic apparatus” (Lee, abstract)
“display, on a display, the data whose quality has been improved by the learned model” (Lee, fig. 1, fig. 11,¶ 0058, ¶ 0073, ¶ 0180: Lee teaches the “display may display information indicating . . . medical information, medical image data, etc.” (¶ 0073) including the improved “image quality of a trained sinogram” (¶ 0180, see also fig. 11).)
“the fourth data is data acquired by performing tomosynthesis imaging of comprising generating X-rays from an X-ray tube and detecting the X-rays that have passed through the subject using an X-ray detector” (Lee, ¶ 0002, ¶ 0019: Lee teaches “generating X-ray tomographic image data” using “tomosynthesis imaging apparatus” (¶ 0002) which includes “an X-ray source configured to radiate X-rays to an object (subject)” and “an X-ray detector configured to detect X-rays radiated by the X-ray source and transmitted by the object” (¶ 0019) where the “X-ray source” discloses an “X-ray tube.”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by explicitly including the type of equipment used to collect the data as explicitly claiming the type of equipment used as disclosed by Lee allows the user more control over system by reducing potential errors to produce improved image quality (Lee, ¶ 0187).
Regarding claim 4 Zhao teaches:
“the fourth data is projection data of the fourth number of views” (Zhao, § III Dataset and Experiments 3rd page: Zhao teaches “We use sample images in the dataset as the original image . . .” (§ III, A. Dataset) where “sample images”… “as the original image” which is based in the “the dataset” which is made up of projection data of the first number of views therefore “the fourth data is projection data of the fourth number of views.”)
“the processing circuitry (see claim 1 above) is further configured to input the fourth data to the learned model to generate processed projection data corresponding to a fifth number of views that is larger than the fourth number of views, and perform reconstruction processing on the generated processed projection data to generate reconstructed image data.” (Zhao, fig. 1, § I. Introduction, 1st page: Zhao teaches using a CNN that outputs the “full-view sinogram” where the “full-view sinogram” (fifth data) is based on the “sparse-view sinogram” (second data) which is based on the “sample images” “as the original image” (fourth data) disclosing the “fifth number of views” is “larger than the fourth number of views.” Moreover, the “full-view sinogram” is reconstructed to generate a reconstructed image (§ I. Introduction).)
Regarding claim 5 Zhao teaches:
“the fourth data is reconstructed image data generated by performing reconstruction processing on projection data of the fourth number of views, and the processing circuitry (see claim 1 above) is further configured to generate reconstructed image data of higher quality than a quality of the fourth data by inputting the fourth data to the learned model” ((Zhao, § III Dataset and Experiments 3rd page: Zhao teaches “We use sample images in the dataset as the original image . . .(§ III, A. Dataset) where the “dataset” is a set of images from a computed tomography, CT, scan (from “Data Science Bowl 2017” (see claim 1 above)) (§ III, A. Dataset)) disclosing the “first data” is “projection data.” Moreover, the images from the CT scan are “reconstructed image data” as a person of ordinary skill in the art would understand the CT scan is not a direct photograph but is created using an algorithm such as iterative reconstruction therefore “second data” is generated on “reconstructed image data reconstructed from the first data.”)
Regarding claim 6 Zhao teaches:
“further comprising a memory that stores, as the learned model, a first learned model that receives input of projection data to improve a quality of the projection data and a second learned model that receives input of reconstructed image data to improve a quality of the reconstructed image data” (Zhao, fig. 1, § III. Dataset and Experiments 3rd page: Zhao teaches “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the CNN discloses a “first learned model” using the “sparse-view sinogram” as input and where the “sparse-view sinogram” is of “projection data.” The CNN outputs the “full-view sinogram” where the “full-view sinogram” is reconstructed to produce a reconstructed image (third data) where the reconstructed image is input for the GAN which outputs the final prediction (fig. 1) disclosing a “second learned model.” where GAN “can significantly reduce the streaking artifacts” (abstract) disclosing improving “a quality of the reconstructed image data.” While Zhao does not explicitly teach memory, Zhao teaches using CNN and GAN which are machine learning techniques in order to manipulate data. A person of ordinary skill in the art would understand machine learning techniques require at least a generic computer which would have memory for storage.)
“wherein the processing circuitry (see claim 1 above) is further configured to input projection data of the fourth number of views to the first learned model to generate processed projection data corresponding to a fifth number of views that is larger than the fourth number of views, performs reconstruction processing on the generated processed projection data to generate processed reconstructed image data, and input the processed reconstructed image data to the second learned model to improve quality of the processed reconstructed image data” (Zhao, fig. 1, Abstract, § II. Method 1st-3rd page: Zhao teaches “The data set used in the experiment is downloaded from the Data Science Bowl 2017. We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) the “sparse-view sinogram” (second data) is based on the “sample images” “as the original image” (fourth data) which comes from a “dataset” (first data) where first data is projection data (see claim 1 above) and therefore fourth data and second data are projection data as they are based in first data and are input into “the first learned model” (CNN) which outputs a “full-view sinogram” which is the fifth data of “a fifth number of views that is larger than the fourth number of views” as the fifth data is a full-view sinogram based on the second data, a sparse-view sinogram, which is based on the fourth data which are sample images. Moreover, Zhao teaches “Reconstruct the images from the completed projection data using the FBP or WHLS methods” (examiner believes WHLS is a typographical error and should read WLS as supported by fig. 1) where the reconstructed images are the images after the FPB/WLS step and before the Estimation Networks (GAN) step of fig 1. The reconstructed image discloses the generated “processed reconstructed image data” which is input into the GAN which produces the Final Prediction (see fig. 1) where GAN, which can “significantly reduce the streaking artifacts” (abstract) discloses the “second learned model to improve quality of the processed reconstructed image data.”)
Regarding claim 9 Zhao teaches:
“the learned model is generated by machine learning using the first data as the third data” (Zhao, § V. Conclusion, 4th and 5th page: Zhao teaches the method “consists of interpolation convolutional neural network in the projection domain and estimation generative adversarial networks in the image domain, which both help to improve the quality of reconstructed images” (§ V. Conclusion) disclosing the “reconstructed images” are a quality representation of the “first data”
Regarding claim 10 Zhao teaches:
“the first data is X-ray data collected by computed tomography” (Zhao, § III Dataset and Experiment, 3rd page: Zhao teaches downloading the data from “Data Science Bowl 2017” (§ III, A. Dataset) which is x-ray data from a CT scan (see claim 1 above).)
Regarding claim 11 Zhao teaches:
“the first data and the second data are X-ray data collected by tomosynthesis imaging” Zhao teaches “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise.” (§ III, A. Dataset) where the “dataset” discloses “first data” and “sample images” “as the original image” discloses the “fourth data” and “the sparse-view. . . sinogram” discloses the “second data” which is based in the “fourth data” which are data from a CT scan.
While Zhao teaches sparse-view data from a CT scan (abstract), Zhao does not teach the data is “collected by tomosynthesis imaging.” Lee teaches “generating X-ray tomographic image data” using “tomosynthesis imaging apparatus” (¶ 0002). Therefore the combination of Zhao and Lee teach the limitation “the first data and the second data are X-ray data collected by tomosynthesis imaging.”
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by explicitly including the X-ray data is collected by tomosynthesis imaging as disclosed by Lee as explicitly claiming the type of equipment used allows the user more control over system by reducing potential errors in order to produce improved image quality (Lee, ¶ 0187).
Regarding claim 12 Zhao teaches:
“the third number of views is smaller than the first number of views” (Zhao teaches “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise.” (§ III, A. Dataset). A reconstructed generated full-view sinogram is the third data with a “third number of views” and is based on the sparse view sinogram (§ I. Introduction, see also fig. 1) which is based on “sample images in the dataset” where the “dataset” is the first view with “the first number of views.” The “sparse-view” sinogram has fewer views than the “dataset” therefore the reconstructed generated full-view sinogram will have fewer views than the “dataset” disclosing “the third number of views is smaller than the first number of views.”)
Regarding claim 13 Zhao teaches;
“the fourth number of views corresponds to the second number of views” (Zhao teaches “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise.” (§ III, A. Dataset) where the “sparse-view” sinogram discloses second data with the “second number of views” where the “sparse-view” sinogram is generated from the “sample images” which discloses the fourth data with the “fourth number of views” thereby disclosing “the fourth number of views corresponds to the second number of views.”)
Regarding claim 14 Zhao teaches:
“the learned model is generated by using a generative adversarial network (GAN) or a discriminator network (DNN)” (Zhao, fig. 3, § II Methods 1st -3rd page: Fig. 3 depicts the GAN Architecture which includes a generator network and a discriminator network (see also § II B. Network Architectures).)
Regarding Independent claim 16
“A medical information processing method” (Zhao teaches processing low dose and sparse view CT (computed tomography) scans (Abstract) disclosing a “medical information processing method.”)
improving a quality of fourth data corresponding to a fourth number of views that is smaller than a first number of views (Zhao, § III Dataset and Experiments, 3rd page: Zhao teaches “We use sample images in the dataset as the original image . . .” (§ III, A. Dataset) where “sample images” “as the original image” discloses the “fourth data corresponding to a fourth number of views” and “the dataset” discloses the “first number of views.” The fourth number of views would be smaller than the first number of views as the “sample images” “as the original image” would be a smaller set that the “dataset” and thereby have a smaller number of views.)
“by inputting the fourth data to a learned model generated by performing machine learning with second data corresponding to a second number of views as input learning data” (Zhao, § III Dataset and Experiments, 3rd page,: Zhao teaches “Train a convolution neural network in the projection domain to complete the data” (§ I. Introduction) and “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the “sparse-view” discloses the “second data corresponding to a second number of views” and where the “sparse-view sinogram and the original full-view sinogram” being used as “the data pair to feed CNN” disclosing “second data” is used as “input learning data.”)
“and third data corresponding to a third number of views that is larger than the second number of views as output learning data” (Zhao, fig 1, § I. Introduction, 1st page: Zhao teaches “Reconstruct the images from the completed projection data using the FBP or WHLS methods” (§ I.) where the reconstructed images are the images after the FPB/WLS step and before the Estimation Networks (AN) step of fig 1. The reconstructed image discloses “third data corresponding to a third number of views.” Moreover, the WHLS (examiner believes this is a typographical error and should read WLS as supported by fig. 1) is part of a machine learning method of which the “third data corresponding to a third number of views” is output. The reconstructed image discloses “third data corresponding to a third number of views” which is “larger than the second number of views” as the “third view” is based on the “full view sinogram” and the “second number of views” is the “sparse-view sinogram” (see above).)
“the second data and the third data being acquired based on first data corresponding that is projection data of the first number of views” (Zhao, fig. 1, § III. Dataset and Experiments 3rd page: Zhao teaches “The data set used in the experiment is downloaded from the Data Science Bowl 2017. We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the CNN outputs the “full-view sinogram” where the “full-view sinogram” is reconstructed to produce a reconstructed image (third data) that is based on the “sparse-view sinogram” (second data) which is based on the “sample images” “as the original image” (fourth data) which comes from a “dataset” therefore is based on “first data.” Zhao teaches the “dataset” is a set of images from a computed tomography, CT, scan (“Data Science Bowl 2017”) as evidenced by Jacobs (see attached “Deep Learning for Lung Cancer Detection on Screening CT Scans” first page) disclosing the “first data” is “projection data.”)
“the second data is data corresponding to a sparser sparse view than the third data and is generated by downsampling the first data” ((Zhao, § III. Dataset and Experiments 3rd page: Zhao teaches using a CNN which inputs a “sparse-view sinogram” (second data) and outputs a “full-view sinogram” which is reconstructed to generate a reconstructed image (third data) disclosing the “second data is data corresponding to a sparser sparse view than the third data.” Additionally, “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view” (§ III, A. Dataset). A person of ordinary skill in the art would understand Matlab, a programming and numeric computing platform that is commonly available, uses “downsampling” which can be used to create a sparse view image as evidenced by Taylor (see attached Matlab Help Center).)
Zhao does not teach:
“displaying, on a display, the data whose quality has been improved by the learned model”
the fourth data is data acquired by performing tomosynthesis imaging comprising generating X-rays from an X-ray tube and detecting the X-rays that have passed through the subject using an X-ray detector.”
Lee teaches:
“displaying, on a display, the data whose quality has been improved by the learned model” (Lee, fig. 1, fig. 11,¶ 0058, ¶ 0073, ¶ 0180: Lee teaches the “display may display information indicating . . . medical information, medical image data, etc.” (¶ 0073) including the improved “image quality of a trained sinogram” (¶ 0180, see also fig. 11).)
“the fourth data is data acquired by performing tomosynthesis imaging comprising generating X-rays from an X-ray tube and detecting the X-rays that have passed through the subject using an X-ray detector” (Lee, ¶ 0002, ¶ 0019: Lee teaches “generating X-ray tomographic image data” using “tomosynthesis imaging apparatus” (¶ 0002) which includes “an X-ray source configured to radiate X-rays to an object (subject)” and “an X-ray detector configured to detect X-rays radiated by the X-ray source and transmitted by the object” (¶ 0019) where the “X-ray source” discloses an “X-ray tube.”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by explicitly including the type of equipment used to collect the data as explicitly claiming the type of equipment used as disclosed by Lee allows the user more control over system by reducing potential errors to produce improved image quality (Lee, ¶ 0187).
Claims 3 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Zhang as modified by Lee, and further in view of Ghani et al., “Integrating Data and Image Domain Deep Learning for Limited Angle Tomography using Consensus Equilibrium,” downloaded from doi: 10.1109/ICCVW.2019.00486.
Regarding claim 3 Zhao teaches:
“the first data is projection data of the first number of views, and the processing circuitry (see claim 1 above) is configured to generate the second data on reconstructed image data reconstructed from the first data” ((Zhao, fig. 2, 2nd page 1st column, 3rd page, § III Dataset and Experiments, A. Dataset: Interpolation network as shown in fig. 2 is part of “processing circuitry.” Zhao teaches “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view . . .” (§ III, A. Dataset) where the “dataset” is a set of images from a computed tomography, CT, scan (from “Data Science Bowl 2017” (§ III, A. Dataset)) disclosing the “first data” is “projection data.” Moreover, the images from the CT scan are “reconstructed image data” as a person of ordinary skill in the art would understand the CT scan is not a direct photograph but is created using an algorithm such as iterative reconstruction therefore “second data” is generated on “reconstructed image data reconstructed from the first data.”)
Zhao does not teach the second data is generated by “performing forward projection.”
Ghani teaches in equations (4) that
y
^
consistent is dependent on
y
l
i
m
i
t
e
d
and A where
y
l
i
m
i
t
e
d
is the “observed limited angle data,” A is “the tomographic forward projection operator,” and
y
^
consistent the “estimated consistent and completed projection data” (second data) (§ 3 1st paragraph).)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by in including performing forward projection as disclosed by Ghani as performing forward projection allows for simulation of how projection data was acquired leading to refining the image based on the measured projections in order to provide a system which can not only “recover lost information.
Regarding claim 18:
Claim 18 recites analogous limitations to claim 3 above and is therefore rejected on the same premise.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Zhang as modified by Lee, and evidenced by Zhang, “The use of artificial intelligence in computed tomography image reconstruction – A literature review” downloaded from https://doi.org/10.1016/j.jmir.2020.09.001.
Regarding claim 7 Zhao teaches:
“the fourth data is projection data of the fourth number of views, and the processing circuitry is further configured to input the fourth data to the learned model to adjust a parameter related to an iterative reconstruction method, and use the adjusted parameter to generate reconstructed image data” (Zhao,fig. 1, § III, A. Dataset: Zhao teaches “The data set used in the experiment is downloaded from the Data Science Bowl 2017. We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where “sample images” “as the original image” discloses the “fourth data” of “the fourth number of views.” The “fourth data” is based on “the dataset” which is a set of images from a computed tomography, CT, scan (“Data Science Bowl 2017”) disclosing “the dataset” is “projection data” therefore the “fourth data” is also “projection data” as it is based on “the dataset.”
While Zhao does not explicitly teach the CNN uses “an iterative reconstruction method” a person of ordinary skill in the art would understand that CNN is a well known deep learning algorithm that uses “an iterative reconstruction method” to adjust parameters to “generate reconstructed image data” as evidenced by Zhang, “The use of artificial intelligence in computed tomography image reconstruction – A literature review” § Deep learning algorithms (DLR) page 675).)
Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Zhao et al., hereinafter Zhao, “Sparse-View CT Reconstruction via Generative Adversarial Networks” downloaded from DOI: 10.1109/NSSMIC.2018.8824362 in view of Lee et al., hereinafter Lee, U.S. Pub. No. 2019/0139276 A1, in view of Bo Zhou et al, “Limited View Tomographic Reconstruction Using a Cascaded Residual Dense Spatial-Channel Attention Network with Projection Data Fidelity Layer” downloaded from doi:10.1109/TMI.2021.3066318 in view of Zhang, “The use of artificial intelligence in computed tomography image reconstruction – A literature review” downloaded from https://doi.org/10.1016/j.jmir.2020.09.001.
Regarding Independent claim 21 Zhao teaches:
“processing circuitry configured to (a) improve a quality of fourth data corresponding to a fourth number of views that is smaller than a first number of views” (Zhao, fig. 2, fig 3, § II. Method 1st – 2nd page, § III Dataset and Experiments, 3rd page: While Zhao does not explicitly teach processing circuitry, Zhao teaches using CNN and GAN which are machine learning techniques in order to manipulate data. A person of ordinary skill in the art would understand machine learning techniques such as CNN and GAN require at least a generic computer which would have processing circuitry which would include the interpolation network (fig. 2) generator and discriminator network (fig. 3). Additionally, Zhao teaches “We use sample images in the dataset as the original image . . .” (§ III, A. Dataset) where “sample images” “as the original image” discloses the “fourth data corresponding to a fourth number of views” and “the dataset” discloses the “first number of views.” The fourth number of views would be smaller than the first number of views as the “sample images” “as the original image” would be a smaller set that the “dataset” and thereby have a smaller number of views.
“by inputting the fourth data to a learned model generated by performing machine learning with second data corresponding to a second number of views as input learning data” (Zhao, § III Dataset and Experiments, 3rd page,: Zhao teaches “Train a convolution neural network in the projection domain to complete the data” (§ I. Introduction) and “We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the “sparse-view” discloses the “second data corresponding to a second number of views” and where the “sparse-view sinogram and the original full-view sinogram” being used as “the data pair to feed CNN” disclosing “second data” is used as “input learning data.”)
“third data corresponding to a third number of views that is larger than the second number of views as output learning data” (Zhao, fig 1, § I. Introduction, 1st page: Zhao teaches “Reconstruct the images from the completed projection data using the FBP or WHLS methods” (§ I.) where the reconstructed images are the images after the FPB/WLS step and before the Estimation Networks (AN) step of fig 1. The reconstructed image discloses “third data corresponding to a third number of views.” Moreover, the WHLS (examiner believes this is a typographical error and should read WLS as supported by fig. 1) is part of a machine learning method of which the “third data corresponding to a third number of views” is output. The reconstructed image discloses “third data corresponding to a third number of views” which is “larger than the second number of views” as the “third view” is based on the “full view sinogram” and the “second number of views” is the “sparse-view sinogram” (see above).)
“the second data and the third data being acquired based on first data corresponding to the first number of views” (Zhao, fig. 1, § III. Dataset and Experiments 3rd page: Zhao teaches “The data set used in the experiment is downloaded from the Data Science Bowl 2017. We use sample images in the dataset as the original image, and simulate the projection process with Matlab to generate the sparse-view and the original full-view sinogram with Poisson noise. The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the CNN outputs the “full-view sinogram” where the “full-view sinogram” is reconstructed to produce a reconstructed image (third data) that is based on the “sparse-view sinogram” (second data) which is based on the “sample images” “as the original image” (fourth data) which comes from a “dataset” therefore is based on “first data.” Zhao teaches the “dataset” is a set of images from a computed tomography, CT, scan (“Data Science Bowl 2017”) as evidenced by Jacobs (see attached “Deep Learning for Lung Cancer Detection on Screening CT Scans” first page) disclosing the “first data” is “projection data.”)
the fourth data is projection data of the fourth number of views” ((Zhao, 3rd page, § III Dataset and Experiments, A. Dataset: Zhao teaches “We use sample images in the dataset as the original image . . .” (§ III, A. Dataset) where “sample images”… “as the original image” which is based in the “the dataset” which is made up of projection data of the first number of views therefore “the fourth data is projection data of the fourth number of views.”)
While Zhao teaches that second data is based fourth data which comes from a dataset and is therefore based on first data. Zhao does not teach “the second data is data corresponding to a limited angle with a narrower angle range than the third data”
Bo Zhou teaches a method of tomographic reconstruction using neural networks for “both limited angle reconstruction and sparse view reconstruction” (abstract) therefore the combination of Zhao and Bo Zhou teach the limitation “the second data is data corresponding to a limited angle with a narrower angle range than the third data.”
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s sparse view CT reconstruction method by incorporating a limited angle rang as disclosed by Bo Zhou in order to achieve “a consistent and substantial improvement over the existing neural network methods” (Bo Zhou, abstract).
Zhao does not teach:
“An X-ray diagnostic apparatus
display, on a display, the data whose quality has been improved by the learned model,
wherein the fourth data is data acquired by performing tomosynthesis imaging comprising generating X-rays from an X-ray tube and detecting the X-rays that have passed through the subject using an X-ray detector”
Lee teaches:
“An X-ray diagnostic apparatus” (Lee, abstract.)
“display, on a display, the data whose quality has been improved by the learned model” (Lee, fig. 1, fig. 11,¶ 0058, ¶ 0073, ¶ 0180: Lee teaches the “display may display information indicating . . . medical information, medical image data, etc.” (¶ 0073) including the improved “image quality of a trained sinogram” (¶ 0180, see also fig. 11).)
“the fourth data is data acquired by performing tomosynthesis imaging comprising generating X-rays from an X-ray tube and detecting the X-rays that have passed through the subject using an X-ray detector” (Lee, ¶ 0002, ¶ 0019: Lee teaches “generating X-ray tomographic image data” using “tomosynthesis imaging apparatus” (¶ 0002) which includes “an X-ray source configured to radiate X-rays to an object (subject)” and “an X-ray detector configured to detect X-rays radiated by the X-ray source and transmitted by the object” (¶ 0019) where the “X-ray source” discloses an “X-ray tube.”)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by explicitly including the type of equipment used to collect the data as explicitly claiming the type of equipment used as disclosed by Lee allows the user more control over system by reducing potential errors to produce improved image quality (Lee, ¶ 0187).
Zhao teaches:
“the processing circuitry (see above) is further configured to input the fourth data to the learned model to adjust a parameter that is at least one of a degree to which the pre-updated data is used and an update speed in an iterative reconstruction method, and use the adjusted parameter to generate reconstructed image data being tomosynthesis image data (Zhao, § III Dataset and Experiments, 3rd page: Zhao teaches “The sparse-view sinogram and the original full-view sinogram is used as the data pair to feed CNN” (§ III, A. Dataset) where the “original full-view sinogram” discloses “fourth data.” The CNN outputs a reconstructed “full-view sinogram.”)
While Zhao teaches using CNN to output a reconstructed sinogram, Zhao does not explicitly teach the CNN uses “an iterative reconstruction method.” A person of ordinary skill in the art would understand that CNN is a well known deep learning algorithm that uses “an iterative reconstruction method” to adjust parameters to “generate reconstructed image data” as taught by Zhang, (§ Deep learning algorithms (DLR) page 675). Additionally, Zhang teaches IR (iterative reconstruction) models are “used to produce estimations of image data, which is compared and corrected relative to the actual obtained data. The correction stage is repeated until difference between estimated and actual data is minimal, or a predetermined threshold is reached” where the “actual data” discloses the “pre-updated data” and the “predetermined threshold is reached” discloses “a degree to which the pre-updated data is used” as the comparisons and corrections are with respect to the “actual data” (pre-updated data). Therefore the combination of Zhao and Zhang teach the limitations “input the fourth data to the learned model to adjust a parameter that is at least one of a degree to which the pre-updated data is used and an update speed in an iterative reconstruction method, and use the adjusted parameter to generate reconstructed image data.”
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by explicitly including how the iterative reconstruction process works as disclosed by Zhang because that knowledge allows for the understanding of how the reconstruction method works and is able to provide for “low dose CT examinations” (Zhang, abstract).
While Zhao teaches using CT (computed tomography) scan data, Zhao does not teach the data is “tomosynthesis image data” Lee teaches “generating X-ray tomographic image data” using “tomosynthesis imaging apparatus” (¶ 0002). Therefore the combination of Zhao, Zhang, and Lee teach the limitations “input the fourth data to the learned model to adjust a parameter that is at least one of a degree to which the pre-updated data is used and an update speed in an iterative reconstruction method, and use the adjusted parameter to generate reconstructed image data being tomosynthesis image data.”
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Zhao’s CT reconstruction method by explicitly including the type of data used such as tomosynthesis data, ensures appropriate statistical analysis is used thereby reducing potential errors in order to produce improved image quality (Lee, ¶ 0187).
Response to Arguments
Applicant’s arguments (remarks) filed on 08/21/2025 have been fully considered.
Regarding Applicant’s arguments (remarks) with respect to the 35 U.S.C. § 103 rejections page 8-13 of Applicant’s remarks, Examiner finds Applicant’s arguments persuasive. New grounds for rejection are presented above.
Regarding Applicant’s arguments (remarks) with respect to the 35 U.S.C. § 101 rejection page 13-17 of Applicant’s remarks, after consideration and consultation, the 35 U.S.C. § 101 rejection has been withdrawn.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Do et al., U.S. Pub. No. 2020/0210767 A1 teaches a method and system for analyzing medical image data using machine learning.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Denise R Karavias whose telephone number is (469)295-9152. The examiner can normally be reached 7:00 - 3:00 M-F.
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, Arleen M. Vazquez can be reached at 571-272-2619. 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.
/DENISE R KARAVIAS/Examiner, Art Unit 2857
/ARLEEN M VAZQUEZ/Supervisory Patent Examiner, Art Unit 2857
Prosecution Insights