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 .
Information Disclosure Statement
The IDS(s) has/have been considered and placed in the application file.
Double Patenting
The provisional rejection of claims 1-7 on the ground of nonstatutory double patenting over claims 1-14 of copending Application No. 18/816,786 is withdrawn in view of the terminal disclaimer filed September 5, 2026, which has been reviewed and accepted.
Response to Arguments
Applicant's arguments filed September 8, 2026, have been fully considered but they are not persuasive.
Applicant's argument that Bonakdar Sakhi discloses only a simple uniform averaging operation and not feature-dependent weighting does not reach the rejection, because Bonakdar Sakhi was never relied on for the feature-dependent selection. Bonakdar Sakhi was relied on for the ensemble of separately trained lesion models and for combining their per voxel probabilities into one integration value, while Gazit was relied on for selecting the coefficient set by the range the measured organ intensity falls in. Bonakdar Sakhi does not limit itself to an unweighted combination, since it states that “The final estimate may be obtained, for example, by any suitable combinational function that evaluates the n estimates, such as an unweighted mean of then estimates or any other suitable combinational function” (¶ 120). The argument attacks the references one at a time and does not reach the combination as claimed.
Applicant relies on the sentence of Gazit beginning “Alternatively, there is a continuum of different sets of organ intensity characteristics” (¶ 93) to argue that Gazit teaches only dynamic on-the-fly coefficient finding, but that sentence is the second of two embodiments given in that paragraph and is introduced by the word “Alternatively.” The sentence preceding it states that “the sets of organ intensity characteristics consist of a finite number of sets, each representing a range of values of the organ intensity characteristics” (¶ 93), and the same paragraph gives the worked example that “training images of a left kidney were grouped into four clusters” (¶ 93). Gazit further states that “each of the sets of organ intensity characteristics is explicitly stored as data in a computer memory or digital storage medium” (¶ 94), which is what the claimed pre-determined coefficients require. The continuum passage applicant relies on is Gazit's alternative embodiment, not Gazit's only teaching. Applicant's argument is therefore unpersuasive.
Applicant's hindsight argument is unpersuasive because the reason to combine is taken from Gazit itself, which states that “The matching set of organ intensity characteristics of each image is chosen, and different values of image processing parameters are used in the image processing, depending on the matching set of organ intensity characteristics chosen for that image” (¶ 90). No teaching was taken from applicant's own disclosure.
Claim Rejections - 35 USC § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-4 and 7 is/are rejected under 35 U.S.C. 103 as being unpatentable over US PG Pub 2022/0138932 A1 to Bonakdar Sakhi et al. (hereinafter "Bonakdar Sakhi") in view of US PG Pub 2016/0300351 A1 to Gazit (hereinafter "Gazit").
Claim 1.
Bonakdar Sakhi and Gazit disclose a lesion detection method for a computer to execute:
a learning process of:
classifying a plurality of first tomographic images obtained by imaging an inside of a plurality of first human bodies into a plurality of first tomographic image groups which have (Bonakdar Sakhi: "This liver/lesion detection stage 130 of the AI pipeline 100 uses an ensemble of ML/DL computer models 132-136 to detect the liver and lesions in the liver as represented in the input volume 105" (¶99). Bonakdar Sakhi trains multiple models on CT tomographic images of human bodies but does not classify training images into groups by medical finding), and
generating a plurality of first lesion identification models for identifying whether or not each unit image region included in a tomographic image as an identification target is a specific lesion region by machine learning which uses (Bonakdar Sakhi: "the ensemble
of ML/DL computer models 132-136 uses differently trained ML/DL computer models 132-
136 to perform liver and lesion detection, with the ML/DL computer models 132-136 being
trained and using loss functions to counterbalance false positives and false negatives in
lesion detection" (¶99). Bonakdar Sakhi generates a plurality of lesion identification
models by machine learning, though the models differ by loss function rather than by
training data group.); and
a lesion detection process of: (Bonakdar
Sakhi does not teach calculating an image feature amount from the test images.), acquiring a probability that each of the unit image regions included in the plurality of second tomographic images is the specific lesion region from each of the plurality of first lesion identification models, by inputting the plurality of second tomographic images to each of the plurality of first lesion identification models (Bonakdar Sakhi: "At run time ... the second ML/DL computer model 620 generates two lesion outputs 624,625" (¶141). "Output probability values range between 0 and 1" (¶137). Each model produces per-voxel lesion probabilities for the input tomographic images), calculating, for each of the unit image regions included in the plurality of second tomographic images, an integration value by performing a weighted addition the probabilities acquired from each of the plurality of first lesion identification models, (Bonakdar Sakhi: "All the generated detections of the ML/DL model 620 for each slab of the input volume 105 are combined with the generated detections of the ML/DL model 630 via the volume averaging (VOL AVG) logic 640. This logic computes the average of the two detection masks at the voxel level." (¶142). Bonakdar Sakhi integrates probabilities from
multiple models at the voxel level but uses uniform averaging rather than weighting based
on an image feature amount.), and detecting the specific lesion region from each of the plurality of second tomographic images based on the integration value (Bonakdar Sakhi: "The result is a Final Lesion mask 650 corresponding to the detected lesions in the input volume 105" (¶142).).
Bonakdar Sakhi discloses all of the subject matter as described above except for specifically teaching (1) classifying training images into groups having different medical findings, (2) calculating an image feature amount from the test images, and (3) selecting a set of predetermined weight coefficients for the weighted addition based on a range in which the
calculated image feature amount falls. However, Gazit teaches that when processing medical images of organs, training images should be clustered by organ intensity characteristics including disease states: (Gazit: "the different sets of organ intensity characteristics correspond, for example, to different results of contrast agent use, or lack of use. Additionally or alternatively, different organ intensity characteristics are due to disease states, such as pneumonia in the lungs or cirrhosis in the liver" (¶90).) Gazit teaches generating separate training data for each clustered group: (Gazit: "for each target organ, the training images are optionally divided into clusters which have different organ intensity characteristics ... A different set of training data is then generated for each cluster" (¶108).) Gazit teaches estimating organ intensity characteristics from the test image: (Gazit: "automatically estimating one or more organ intensity characteristics in the image, from contents of a region of the image that appears to correspond, at least in part, to at least a portion of the organ" (claim 1).) And Gazit teaches selecting one of a finite number of pre-stored sets according to
the range in which the estimated characteristic falls: (Gazit: “the sets of organ intensity characteristics consist of a finite number of sets, each representing a range of values of the organ intensity characteristics ... An example of such sets is shown below, in FIG. 2, where training images of a left kidney were grouped into four clusters, each cluster containing images with similar results of contrast agent use or lack of use, and each cluster is used to generate one of the four sets of organ intensity characteristics” (¶ 93). “each of the sets of organ intensity characteristics is explicitly stored as data in a computer memory or digital storage medium” (¶ 94). “The matching set of organ intensity characteristics of each image is
chosen, and different values of image processing parameters are used in the image
processing, depending on the matching set of organ intensity characteristics chosen for that
image” (¶ 90).)
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify Bonakdar Sakhi's ensemble lesion detection system, which uses uniform volume
averaging to combine model outputs without accounting for patient-specific imaging
conditions, to incorporate Gazit's clustering and range-based selection of pre-stored coefficient
sets. Gazit teaches that organ intensity varies by disease state and that this variability should drive how image processing adapts to the test patient. A person of ordinary skill would recognize that modifying Bonakdar Sakhi 's averaging to weight model outputs based on estimated organ characteristics would improve detection accuracy across patients with varying organ conditions such as steatosis or cirrhosis.
Claim 2.
Bonakdar Sakhi and Gazit disclose the lesion detection method according to claim 1, wherein the first image feature amount is calculated based on pixel information on a region of a specific organ among image regions of the plurality of second tomographic images (Gazit: "organ intensity characteristics mean intensity characteristics of a region of the image that appears to correspond, at least in part, to at least a portion of the organ" (¶89).).
Claim 3.
Bonakdar Sakhi and Gazit disclose the lesion detection method according to claim 2, wherein the first image feature amount is an average luminance value in the region of the specific organ (Gazit: "the intensity characteristics may comprise ... a mean, median, standard deviation, or other moment of any of these distributions" (¶89). A mean of the intensity distribution over
the organ region is an average luminance value.).
Claim 4.
Bonakdar Sakhi and Gazit disclose the lesion detection method according to claim 3, wherein the plurality of first tomographic image groups are classified based on an accumulation degree of fat in the specific organ (Gazit: “The different sets of organ intensity characteristics
correspond, for example, to different results of contrast agent use, or lack of use. Additionally or alternatively, different organ intensity characteristics are due to disease states, such as pneumonia in the lungs or cirrhosis in the liver” (¶90). Fat accumulation in the liver is a disease state that alters organ intensity characteristics within Gazit’s clustering framework.), and in the calculating of the integration value, by performing weighting addition on the probability acquired from each of the plurality of first lesion identification models, the integration value is calculated (Gazit: “a different linear combination of a finite number of basis sets of organ intensity characteristics … finding coefficients of the basis sets that would lead to a linear combination” (¶93). A linear combination with coefficients is weighting addition), and as the average luminance value is lower, a higher weight coefficient is set for the probability from a first lesion identification model generated by using the first tomographic image group which has a higher accumulation degree of fat, among the plurality of first lesion identification models (Gazit's: the coefficients are found to make the “best match … to the organ intensity characteristics of the image (¶93), and the organ intensity characteristics include a “a mean … of these distributions” (¶89). Gazit’s list of disease states is non-exhaustive (“such as … cirrhosis in the liver … or due to other individual differences” ¶90), encompassing liver condition that alters organ intensity characteristics, including hepatic steatosis (fat accumulation). Because Gazit’s coefficient-finding selects coefficients that best match the test image’s estimated organ characteristic to the basis sets, when the test image has lower average luminance – corresponding to higher fat accumulation – the coefficients inherently assign higher weight to the basis set derived from the higher-fat-accumulation cluster, as that cluster’s intensity characteristics are the closer match.).
Claim 7.
Regarding claim 7, claim 7 recites a non-transitory computer-readable recording medium storing a lesion detection program causing a computer to execute the same method steps as claim 1. Bonakdar Sakhi and Gazit disclose all limitations for the same reasons as claim 1. (Bonakdar Sakhi claim 11: "A computer program product comprising a computer readable storage medium having a computer readable program stored therein ... causes the computing device to implement a lesion detection ensemble machine learning model architecture.").
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Bonakdar Sakhi in view of Gazit as applied to claim 1 above, and further in view of Gyorfi et al. “Brain Tumor Detection and Segmentation from Magnetic Resonance Image Data Using Ensemble Learning Methods” (hereinafter "Gyorfi").
Claim 5.
Bonakdar Sakhi and Gazit disclose the lesion detection method according to claim 1, wherein the learning process includes a process of: calculating, for each second tomographic image group into which the plurality of first tomographic images are classified for each of the first human bodies, a second image feature amount of a same type as the first image feature amount (Gazit: "automatically estimating one or more organ intensity characteristics in the image, from contents of a region of the image that appears to correspond, at least in part, to at least a portion of the organ" (claim 1). Gazit calculates organ intensity characteristics for each training image; the same type of feature is computed for each image in each finding-classified group.), classifying the plurality of first tomographic images into a plurality of third tomographic image groups according to a range of the second image feature amount (Gazit: “the sets of organ intensity characteristics consist of a finite number of sets, each representing a range of values of the organ intensity characteristics, for example based on a cluster of training images that has similar but not identical organ intensity characteristics” (¶93). Gazit teaches partitioning training images into sub-clusters according to ranges of the intensity characteristics.), and generating ones different from each other among the plurality of third tomographic image groups as learning data (Gazit: “for each target organ, the training images are optionally divided into clusters which have different organ intensity characteristics … A different set of training data is then generated for each cluster, for example the target organ may have different distributions of intensity values and gradients in the training images of different clusters” (¶108). Bonakdar Sakhi: " The ensemble of ML/DL computer models 132-136 uses differently trained ML/DL computer models 132-136 to perform liver and lesion detection" (¶99).), the lesion detection process includes a process of: acquiring the probability for each of the unit image regions included in the plurality of second tomographic images from each of the plurality of second lesion identification models, by inputting the plurality of second tomographic images to each of the plurality of second lesion identification models (Bonakdar Sakhi: "At run time ... the second ML/DL computer model 620 generates two lesion outputs 624,625" (¶141). "Output probability values range between 0 and 1" (¶137). The second set of models, like the first, produces per-pixel lesion probabilities for the input tomographic images.), and in the calculating of the integration value, and for each of the unit image regions included in the plurality of second tomographic images, the integration value is calculated by integrating the probabilities acquired from each of the plurality of first lesion identification models and each of the plurality of second lesion identification models, based on the first image feature amount (Gazit: “a different linear combination of a finite number of basis sets of organ intensity characteristics … finding coefficients of the basis sets that would lead to a linear combination” (¶93). Bonakdar Sakhi: "All the generated detections of the ML/DL model 620 for each slab of the input volume 105 are combined with the generated detections of the ML/DL model 630 via the volume averaging (VOL AVG) logic 640. This logic computes the average of the two detection masks at the voxel level" (¶142). Gazit’s coefficient-based linear combination applied to Bonakdar Sakhi’s voxel-level integration of multiple model outputs teaches integrating probabilities from both first and second sets of models based on the first image feature amount.).
Bonakdar Sakhi and Gazit discloses all of the subject matter as described above except for specifically teaching a plurality of second lesion identification models in addition to the first set. However, Gyorfi in the same field of endeavor teaches a plurality of second lesion identification models (Gyorfi: “13 further features are generated, which were selected in a previous study [29] out of 100 generated morphological, gradient, and Gabor features” (p. 910, Section II-A); Gyorfi teaches generating diverse ensembles of classifiers from different feature-space partitions of the same data the combination teaches generating a second set of lesion identification models from the sub-classified training image groups).
Therefore, it would have been obvious to one of ordinary skill in the art to combine Bonakdar Sakhi, Gazit, and Gyorfi before the effective filing date of the claimed invention. The motivation for this combination of references would have been to improve lesion detection accuracy by capturing fine grain variation within each first-level cluster group, as Gyorfi demonstrates that “segmentation quality rises together with the size of the ensemble up to 125 units” (p. 911, Section III).
Allowable Subject Matter
Claim 6 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
THIS ACTION IS MADE FINAL. 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 Ross Varndell whose telephone number is (571)270-1922. The examiner can normally be reached M-F, 9-5 EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, O’Neal Mistry can be reached at (313)446-4912. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see https://ppair-my.uspto.gov/pair/PrivatePair. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Ross Varndell/Primary Examiner, Art Unit 2674
Prosecution Insights