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 .
Response to Arguments
In response to the amendments and remarks, filed by Applicant on 07/08/2026, the claim objections have been withdrawn.
Applicant’s arguments, see remarks, filed 07/08/2026, with respect to the rejection of claim 10-11 and 14 under 35 U.S.C. 102 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 Therefore, the rejection has been withdrawn.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 Chinese Patent Application Publication No.: CN 101292871 A (Tian et al.) under 35 U.S.C. 103.
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 10-11 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over non-patent literature "Localized Energy-Based Normalization of Medical Images: Application to Chest Radiography"; IEEE Transactions on Medical Imaging, vol. 34, no. 9, pp. 1965-1975, Sept. 2015, doi: 10.1109/TMI.2015.2418031 (Philipsen et al.) (hereinafter Philipsen), in view of Chinese Patent Application Publication No.: CN 101292871 A (Tian et al.) (hereinafter Tian).
Regarding claim 10, Philipsen teaches a computer-implemented method for training a medical image analysis algorithm, the method comprising: (Philipsen, abstract: “Automated quantitative analysis systems for medical images often lack the capability to successfully process images from multiple sources. Normalization of such images prior to further analysis is a possible solution to this limitation. This work presents a general method to normalize medical images and thoroughly investigates its effectiveness for chest radiography (CXR) … We conclude that the normalization can be successfully applied in chest radiography and makes supervised systems more generally applicable to data from different sources”)
receiving a medical image of a body part of a patient (Philipsen, page 1967, FIG. 1 and FIG. 1:
PNG
media_image1.png
350
1020
media_image1.png
Greyscale
;
“Fig. 1. Two images from each of the six datasets are displayed. The top row contains normal CXRs and the bottom row contains abnormal CXRs. From left to right the images come from: odelcadr, atomed, philips, siemens, odelcadr-kv and JSRT. Except for the JSRT images, all images are shown using original DICOM window center/width settings”)
generating a modified medical image (Philipsen, abstract; page 1968, Section IV. Evaluation Design, para. 2; FIG. 3: “The method starts with an energy decomposition of the image in different bands. Next, each band's localized energy is scaled to a reference value and the image is reconstructed.”; “A set of R=50 images from the same scanner as dataset B were selected as reference images as this was the main source of data. All images in this set were normal images. An initial set of pilot experiments, with varying B in the range of 3 to 9, was conducted to determine the optimal number of frequency bands and we chose B=6 as the optimal number of bands. Fig. 3 shows an example of the frequency decomposition”;
PNG
media_image2.png
248
1024
media_image2.png
Greyscale
; and
training a medical image analysis machine learning algorithm utilizing the modified medical image (Philipsen, page 1969, Section 3) Results; FIG. 6; page 1968, para. 1; FIG. 2: “For each method and for each training set, the overlap results of all images are shown in Fig. 6. The figure is divided in six columns with four boxplots: one column for each source of training. The blue, red, magenta and green boxes represent the results of the baseline method, histogram equalization, normalization with one iteration and the iterated normalization, respectively. Table II shows all values of Pk(di,dj) and it's variation. Less brightly colored cells in the columns implies that the segmentation performance is less dependent on the utilized training set. From the table and boxplots it is seen that the segmentation performance has improved compared to the baseline and histogram equalization when the proposed normalization method is used.”;
PNG
media_image3.png
98
508
media_image3.png
Greyscale
;
PNG
media_image4.png
508
506
media_image4.png
Greyscale
;
“After this first stage a rough lung segmentation can be obtained from a supervised method [2], and
the resulting lung region was set as the region of interest in the second stage to give optimal lung standardization. The process is summarized in the flowchart in Fig. 2.”;
PNG
media_image5.png
148
498
media_image5.png
Greyscale
).
wherein the generating comprises a modification of two or more spatial frequency bands associated with the medical image (Philipsen, abstract; page 1968, Section IV. Evaluation Design, para. 2; FIG. 3; see rejection above in the step of generating a modified medical image).
Philipsen fails to teach
wherein the generating comprises a modification of two or more spatial frequency bands associated with the medical image with a random modification factor.
Tian teaches
wherein the generating comprises a modification of two or more spatial frequency bands associated with the medical image with a random modification factor (Tian, page 7, para. 7; page 8, para. 1: “five shapes and sizes are not the same of the simulated active region (FIG. 6a), the number of voxels contained respectively are 10, 30 and 90,180,270. in the active region, each stimulus space effect caused by high AWGN noise distribution. the outside of the activation region, the stimulation does not cause effect. Gaussian noise simulating various activation caused by stimulation mode can make the effect signal energy is uniformly distributed on different spatial frequency band, so the simulation activation brain areas of the local average signal (low frequency component) and a fine space structure (high-frequency component) contains stimulus condition related information, is closer to the actual fMRI data. Then, the stimulus caused by brain activity signal is added to the spatiotemporal noise background. In order to simulate the real noise in the fMRI data local spatial correlation characteristics, background noise produced in the following way: firstly, simulating and obtaining space Gaussian white noise, then using the FWHM (Full Width at Half Maximum, FWHM) of 3.5 Gauss generated white Gaussian noise is smoothed, obtaining the noise background of the local spatial correlation characteristic. for checking activation area extraction algorithm has different contrast-to-noise ratio at different contrast noise ratio (Ccmtrast to Noise Ratio, CNR) performance, five sets of data is generated, the value ratio noise ratio respectively is 0.2, 0.4, 0.6, 0.8, 1.0. comparing the noise ratio is defined as asarone regions in space of each of the average signal amplitude of the maximum absolute value of the background noise standard difference ratio. A.2 simulation data analysis analyzing the simulation data by the following three ways: (1) original data -GLM analysis, (2) Gaussian kernel smoothing analysis, data -GLM (3) LMDM analysis.”; Gaussian noise acts as a random modification factor across spatial frequency bands; it adds independent random values to image pixels; this process spreads flat, uniform changes across all spatial frequencies in the Fourier domain).
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the step of generating including a modification of two or more spatial frequency bands associated with the medical image, as taught by Philipsen, to include being done with a random modification factor, as taught by Tian.
The suggestion/motivation for doing so would have been that adding Gaussian noise to modify spatial frequency bands in medical images helps improve algorithm robustness, test model limits under stress, and protect patient privacy by preventing deep feature extraction.
Therefore, it would have been obvious to combine Philipsen, with Tian, to obtain the invention as specified in claim 10.
Regarding claim 14, Philipsen teaches a computer-implemented medical image analysis method, comprising: (Philipsen, abstract: “Automated quantitative analysis systems for medical images often lack the capability to successfully process images from multiple sources. Normalization of such images prior to further analysis is a possible solution to this limitation. This work presents a general method to normalize medical images and thoroughly investigates its effectiveness for chest radiography (CXR) … We conclude that the normalization can be successfully applied in chest radiography and makes supervised systems more generally applicable to data from different sources”)
receiving an examination medical image of a body part of a patient (Philipsen, page 1967, FIG. 1:
PNG
media_image1.png
350
1020
media_image1.png
Greyscale
;
“Fig. 1. Two images from each of the six datasets are displayed.”);
analyzing the body part of the patient, wherein the analyzing comprises interrogating the examination medical image with a trained machine learning algorithm (Philipsen page 1968, para. 1; FIG. 2: “After this first stage a rough lung segmentation can be obtained from a supervised method [2], and the resulting lung region was set as the region of interest in the second stage to give optimal lung standardization. The process is summarized in the flowchart in Fig. 2.”;
PNG
media_image5.png
148
498
media_image5.png
Greyscale
);
wherein the machine learning algorithm was trained by: receiving the medical image of a body part of a patient; generating a modified medical image (Philipsen, abstract; page 1968, Section IV. Evaluation Design, para. 2; FIG. 3: “The method starts with an energy decomposition of the image in different bands. Next, each band's localized energy is scaled to a reference value and the image is reconstructed.”; “A set of R=50 images from the same scanner as dataset B were selected as reference images as this was the main source of data. All images in this set were normal images. An initial set of pilot experiments, with varying B in the range of 3 to 9, was conducted to determine the optimal number of frequency bands and we chose B=6 as the optimal number of bands. Fig. 3 shows an example of the frequency decomposition”;
PNG
media_image2.png
248
1024
media_image2.png
Greyscale
); and
training the machine learning algorithm utilizing the modified medical image, wherein the generating comprises a modification of two or more spatial frequency bands associated with the medical image (Philipsen, abstract; page 1968, Section IV. Evaluation Design, para. 2; FIG. 3; see above discussion different spatial frequency band modification; page 1969, Section 3) Results; FIG. 6; page 1968, para. 1; FIG. 2: “For each method and for each training set, the overlap results of all images are shown in Fig. 6. The figure is divided in six columns with four boxplots: one column for each source of training. The blue, red, magenta and green boxes represent the results of the baseline method, histogram equalization, normalization with one iteration and the iterated normalization, respectively. Table II shows all values of Pk(di,dj) and it's variation. Less brightly colored cells in the columns implies that the segmentation performance is less dependent on the utilized training set. From the table and boxplots it is seen that the segmentation performance has improved compared to the baseline and histogram equalization when the proposed normalization method is used.”;
PNG
media_image3.png
98
508
media_image3.png
Greyscale
;
PNG
media_image4.png
508
506
media_image4.png
Greyscale
;
“After this first stage a rough lung segmentation can be obtained from a supervised method [2], and
the resulting lung region was set as the region of interest in the second stage to give optimal lung standardization. The process is summarized in the flowchart in Fig. 2.”;
PNG
media_image5.png
148
498
media_image5.png
Greyscale
).
Philipsen fails to teach
wherein the generating comprises a modification of two or more spatial frequency bands associated with the medical image with a random modification factor.
Tian teaches
wherein the generating comprises a modification of two or more spatial frequency bands associated with the medical image with a random modification factor (Tian, page 7, para. 7; page 8, para. 1: “five shapes and sizes are not the same of the simulated active region (FIG. 6a), the number of voxels contained respectively are 10, 30 and 90,180,270. in the active region, each stimulus space effect caused by high AWGN noise distribution. the outside of the activation region, the stimulation does not cause effect. Gaussian noise simulating various activation caused by stimulation mode can make the effect signal energy is uniformly distributed on different spatial frequency band, so the simulation activation brain areas of the local average signal (low frequency component) and a fine space structure (high-frequency component) contains stimulus condition related information, is closer to the actual fMRI data. Then, the stimulus caused by brain activity signal is added to the spatiotemporal noise background. In order to simulate the real noise in the fMRI data local spatial correlation characteristics, background noise produced in the following way: firstly, simulating and obtaining space Gaussian white noise, then using the FWHM (Full Width at Half Maximum, FWHM) of 3.5 Gauss generated white Gaussian noise is smoothed, obtaining the noise background of the local spatial correlation characteristic. for checking activation area extraction algorithm has different contrast-to-noise ratio at different contrast noise ratio (Ccmtrast to Noise Ratio, CNR) performance, five sets of data is generated, the value ratio noise ratio respectively is 0.2, 0.4, 0.6, 0.8, 1.0. comparing the noise ratio is defined as asarone regions in space of each of the average signal amplitude of the maximum absolute value of the background noise standard difference ratio. A.2 simulation data analysis analyzing the simulation data by the following three ways: (1) original data -GLM analysis, (2) Gaussian kernel smoothing analysis, data -GLM (3) LMDM analysis.”; Gaussian noise acts as a random modification factor across spatial frequency bands; it adds independent random values to image pixels; this process spreads flat, uniform changes across all spatial frequencies in the Fourier domain).
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to modify the step of generating including a modification of two or more spatial frequency bands associated with the medical image, as taught by Philipsen, to include being done with a random modification factor, as taught by Tian.
The suggestion/motivation for doing so would have been that adding Gaussian noise to modify spatial frequency bands in medical images helps improve algorithm robustness, test model limits under stress, and protect patient privacy by preventing deep feature extraction.
Therefore, it would have been obvious to combine Philipsen, with Tian, to obtain the invention as specified in claim 14.
Regarding claim 11, Philipsen, in view of Tian, teaches the method according to claim 10, further comprising:
generating a scaled image set comprising a plurality of scaled images, wherein the generating comprises utilizing the medical image, wherein each scaled image comprises a representation of a spatial frequency in the medical image, and wherein the representation of the spatial frequency in each of the plurality of scaled images is different; generating a modified scaled image set from the scaled image set, wherein the generating comprises modifying two or more scaled images of the plurality of scaled images; and generating the modified medical image comprising utilizing the modified scaled image set (Philipsen, abstract; page 1968, Section IV. Evaluation Design, para. 2; FIG. 3; see rejection of claim 10 above showing the different modified images with different frequency bands that are all scaled; page 1971, Section VI. Discussion, para. 2; page 1974, Conclusion: “Two key elements in the proposed method are the applied energy band scaling and the addition of a region of interest … The decomposition into frequency bands provides a separation of structures of different sizes, which is a useful property, as it allows for applying specific scaling factors to each band, which can enhance or suppress specific structures … By taking all λi(Ω) values equal to a reference value, the image's frequency information is standardized among images, which gives them similar appearance and intensity characteristics.”; “The method uses an energy decomposition of the image, after which the energy of each band is scaled separately to a reference energy to acquire a normalized image.”).
Allowable Subject Matter
Claim 16 is allowed. The following is an examiner’s statement of reasons for allowance: with respect to the pending claim, the Prior Art of Record fails to teach, disclose or render obvious the applicant's invention as claimed. The most similar invention in the prior art of record to the claimed invention of independent claim 16 is Philipsen.
Philipsen fails to teach the following limitations from independent claim 16:
“wherein: the modification of the two or more of the plurality of scaled images comprises modification of the representation of the spatial frequency comprised within each of the two or more scaled images with a random modification factor; and/or
the generation of the modified scaled image set comprises a modification of the plurality of scaled images, wherein the modification of the plurality of scaled images comprises a modification of the representation of the spatial frequency comprised within each of the plurality scaled images with: a modification factor that increases with spatial frequency from one scaled image to the next scaled image; or a modification factor that decreases with spatial frequency from one scaled image to the next scaled image”.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: non-patent literature “How to Improve Deep Learning Model Robustness by Adding Noise”; Machine Learning Mastery; Deep Learning Performance; 08/28/2020; https://machinelearningmastery.com/how-to-improve-deep-learning-model-robustness-by-adding-noise/ (Brownlee). Brownlee teaches that noise can be added to a neural network model via the Gaussian Noise layer and explains details of how this happens; this is an example of a adding a “random modification factor” as recited in claim 1, to spatial frequency bands in an image; Gaussian noise is flat across all spatial frequencies (white noise), meaning it uniformly impacts low, mid, and high spatial frequency bands. When added to an image or signal, it injects random variations that corrupt fine edges and macro structures alike, making high-frequency details harder to distinguish from static.
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 MICHAEL ADAM SHARIFF whose telephone number is 571-272-9741. The examiner can normally be reached M-F 8:30-5PM.
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, Sumati Lefkowitz can be reached on 571-272-3638. 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.
/MICHAEL ADAM SHARIFF/
Examiner, Art Unit 2672
/SUMATI LEFKOWITZ/Supervisory Patent Examiner, Art Unit 2672
Prosecution Insights