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 .
Election/Restrictions
Applicant's election with traverse of Group I (claims 1-11) in the reply filed on 2/20/2026 is acknowledged. The traversal is on the ground(s) that “Claim 1 has been amended to recite "preprocessing, by at least one computing device comprising processing circuitry including a processor and memory, activation data... and determining, by the at least one computing device, an anatomical location..." Accordingly, Applicant respectfully submits that unity of invention exists between groups I and II.” Examiner respectfully agrees with the Applicant and the prior election/restriction requirement has been withdrawn. Accordingly, claims 1-20 remain pending for examination.
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(s) 1-20 is/are 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. Claims 2-9 and 11-20 are also rejected at least by virtue of dependency upon a rejected base claim.
Claims 1 and 10 recite the limitations “activation data obtained from brains of multiple subjects to generate one or more dynamic parcellated supervoxel maps based on a whole brain map of the multiple subjects,” which render the claims indefinite. The use of the singular ‘whole brain map’ is unclear when referring to the multiple subjects, because the activation data is obtained from multiple subject brains and is used to generate a plurality of ‘dynamic parcellated supervoxel maps’. The claim language is not consistent with the number and type of data when referring to the multiple subjects. The limitations must be amended to clearly indicate how many ‘maps’ are generated and to clearly define what each ‘map’ is. For the purposes of examination, the broadest reasonable interpretation of one or more ‘whole brain map’ reads on the limitations.
Claims 1 and 10 further recite the limitations “an anatomical location of the functional task in a brain of another subject” which render the claims indefinite. There is insufficient antecedent basis for this limitation in the claim, because it is unclear if ‘another subject’ may be one of the subjects of the ‘multiple subjects’ or may be a different and distinct subject. It is suggested to amend the preamble of the claim to clearly point out the ‘subject’ whose functional tasks are predicted by the method, and to distinguish from the patients whose imaging data contribute the activation data used for generating parcellated supervoxel maps.
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.
Claim(s) 1-4, 7-13 and 16-20 is/are rejected under 35 U.S.C. 103 as being obvious over Huo et al. (“Supervoxel based method for multi-atlas segmentation of brain MR images”, NeuroImage 175 (2018), 201-214, ISSN 1053-8119, 2018-04-10; hereinafter “Huo”), in view of Parisot et al. (“Group-wise parcellation of the cortex through multi-scale spectral clustering”, NeuroImage, v.136 (2016), p.68-83, ISSN 1053-8119, 2016-06-17; hereinafter “Parisot”).
Regarding claim 1, Huo teaches a method for dynamic supervoxel parcellation (“The paper proposes a new segmentation framework based on supervoxels” [abst]; “multi-atlas segmentation describes the parcellation procedure through a probabilistic model constructed on a limited number of manually delineated atlas images” [p.201]; [p.202-206], [fig. 1-4, 6, 8, 10-11]), comprising:
preprocessing, by at least one computing device comprising processing circuitry including a processor and memory, data obtained from brains of multiple subjects to generate one or more dynamic parcellated supervoxel maps based on a whole brain map of the multiple subjects (“Supervoxel segmentation is the first step in the supervoxel graph construction” [p.203]; “Moreover, to calculate the data term, each supervoxel in the atlases should have the unique label. Therefore, a refinement scheme is proposed following the supervoxel segmentation of the atlases.” [p.204]; “The proposed approach is thoroughly evaluated on three publically available brain datasets for the whole brain segmentation” [p.206]; “Before applying our approach, we performed the pre-processing steps in the following order for all the tests: bias field correction, pairwise registration, and image normalization. […] In the multi-atlas segmentation, both the intensity images and the label images of the atlases are required to be warped to the target domain with the same transformation. […] 1) The standard scale landmarks corresponding to each decile were calculated using the warped atlas intensity images; and 2) the image histogram (target and atlas intensity images) was mapped to the standard scale landmarks in a piece-wise linear fashion.” [p.207]; “The average training time for the SVM classifier is 2 h on a single desktop PC (i7-4900 CPU 3.60 GHz, 16 GB RAM).” [p.212]; [p.203-209], [fig. 1-4, 6, 8, 10-11]); and
determining, by the at least one computing device, an anatomical location in a brain of another subject based upon classification of supervoxels of the one or more dynamic parcellated supervoxel maps (“The SVM classifier is used to generate the predicted label image of the target for supervoxel segmentation and the probability map for initialization of data term in dense labeling.” [p.203, fig. 1 inset]; “In the implementation of the supervoxel segmentation, voxel feature vectors randomly selected from the unregistered atlas intensity images are used as the training samples to train the SVM classifier while the voxel feature vectors extracted from each voxel in the target intensity image are treated as the testing samples. […] By applying the SVM classifier to the testing samples, a predicted label image and a probability map of the target are generated simultaneously.” [p.205]; “Unlike the other baseline methods, the result of the proposed method does not contain the undesired “holes” in the anatomical structures, which indicates the advantages of our approach in maintaining the spatial consistency […] We randomly selected 9 subjects as the atlases and the remaining as the test subjects. The SVM classifier with c ¼ 22 and γ ¼ 20 is used to obtain the predicted label image” [p.211]; “The observed qualitative performance improvement demonstrates that the introduction of supervoxels contributes to preserving the spatial label consistency. As shown in the MICCAI 2012 dataset with 134 anatomical structures, the proposed method highlights its benefits in maintaining the spatial label consistency, particularly for the small structures.” [p.212]; The SVM classifier is trained to perform segmentation on multiple atlases and applied to individual training subjects to test the efficacy of the trained model [p.205-213], [fig. 1-4, 6, 8, 10-11]);
but Huo may fail to teach activation data associated with a functional task.
However, in the same field of endeavor, Parisot teaches a method for functional task prediction with dynamic supervoxel parcellation (“In this paper, we propose a group-wise connectivity-driven parcellation method based on spectral clustering that captures local connectivity information at multiple scales and directly enforces correspondences between subjects.” [abst]; “We present an extended experimental evaluation of the parcellation method, notably through comparisons to cytoarchitectonic and fMRI data.” [p.69]; “We evaluate how well the parcellations agree with the underlying brain structure by comparing to task fMRI activation maps.” [p.74]; [p.68-72], [fig. 1-4, 8, 11-13]), comprising:
preprocessing activation data obtained from brains of multiple subjects to generate one or more dynamic parcellated supervoxel maps based on a whole brain map of the multiple subjects, the activation data associated with a functional task (“The structural and diffusion data have been preprocessed following the HCP's minimal preprocessing pipelines” [p.70]; “The task fMRI data is preprocessed following the HCP preprocessing pipelines (gradient unwarping, motion and distortion correction, registration to the MNI space and projection to the cortical surface). Task activation maps are then obtained using standard FSL tools (FEAT) that use general linear modelling to construct activation maps. […] The analysis is carried out across sessions (single subject activation maps) and then across subjects (group-wise activation map)” [p.76]; [p.69-72], [fig. 1-4, 8, 11-13]); and
determining an anatomical location of the functional task in a brain based upon classification of supervoxels of the one or more dynamic parcellated supervoxel maps (“we observe strong correspondences between our parcels' boundaries and strong variations of myelination, specifically in the motor areas, both on the average map and the single subject level […] On average over all parcellation resolutions, we obtain very good overlap measurements with the motor areas (BA 1–6),” [p.77]; “At the single subject level, this allows to estimate which brain regions are the most consistent (inter-subject variability), while the group level enables to evaluate the fundamental differences in connectivity and function between two different groups.” [p.81]; [p.77-83], [fig. 1-4, 8, 11-13; see fig. 13 reproduced below]).
PNG
media_image1.png
800
824
media_image1.png
Greyscale
The anatomical location in the brain corresponding with a functional task supervoxel is determined by the parcellation method (Parisot [fig. 13])
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The state-of-the-art techniques in multi-atlas segmentation, including both patch-based and learning-based methods, are strongly dependent on the pairwise registration or exhibit huge spatial inconsistency (Huo [abst.]). One of the main advantages of using a group-wise parcellation method is the possibility to perform direct comparisons between subjects as well as groups (gender, age or diseased base groups). At the single subject level, this allows to estimate which brain regions are the most consistent (inter-subject variability), while the group level enables to evaluate the fundamental differences in connectivity and function between two different groups (Parisot [p.81]). The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 2, Huo and Parisot teach the method of claim 1,
Huo further teaching wherein the preprocessing comprises:
registering and averaging the activation data of the multiple subjects to produce the whole brain map (“Before implementing the proposed method, pairwise registrations are performed between the target and each atlas. Then, in order to construct the supervoxel graph, supervoxel segmentations are applied on the target and atlases, respectively” [p.203]; [p.205-213], [fig. 1-4, 6, 8, 10-11]); and
generating the one or more dynamic parcellated supervoxel maps from the whole brain map (“a dense labeling step is proposed to acquire the refined label map of the target” [p.203]; “to construct the matching pairs, we randomly select equal supervoxel sample pairs per class from the registered atlases. Then the equal number of non-matching pairs are randomly selected” [p.206]; [p.205-213], [fig. 1-4, 6, 8, 10-11]).
Regarding claim 3, Huo and Parisot teach the method of claim 2,
Huo further teaching wherein supervoxels of the one or more dynamic parcellated supervoxel maps are identified from the whole brain map (“The SVM classifier is used to generate the predicted label image of the target for supervoxel segmentation and the probability map for initialization of data term in dense labeling.” [p.203, fig.1 inset]; Labels are predicted to identify the supervoxels [p.205-213], [fig. 1-4, 6, 8, 10-11], [see claim 1 rejection]).
Regarding claim 4, Huo and Parisot teach the method of claim 3,
Parisot further teaching wherein the supervoxels are identified using a 1-way ANOVA analysis (“We also visualise the average intra and inter-subjects variance parameters of the model (σ1 and σ2 respectively) […] the intra-subject variance monotonously decreases with the number of parcels, while the inter-subject variance follows the opposite trend.” [p.79]; [p.77-83], [fig. 1-4, 8, 11-13, 18; see fig. 18 reproduced below]).
PNG
media_image2.png
628
748
media_image2.png
Greyscale
Analysis of variance between group-wise parcellations of supervoxels, both Intra and inter-subject (Parisot [fig. 18])
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 7, Huo and Parisot teach the method of claim 1,
Parisot further teaching wherein the functional task is associated with a foot, a hand, or a mouth of the subject (“For all resolutions, we observe strong correspondences between our parcels' boundaries and strong variations of myelination, specifically in the motor areas, both on the average map and the single subject level.” [p.77]; [p.77-83], [fig. 1-4, 8, 11-13; see fig. 13 reproduced below]).
PNG
media_image1.png
800
824
media_image1.png
Greyscale
Left Hand and Tongue motor tasks associated with parcellated supervoxel labels (Parisot [fig. 13])
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 8, Huo and Parisot teach the method of claim 1,
Parisot further teaching wherein the activation data is acquired through magnetic resonance imaging of the subject (“We present an extended experimental evaluation of the parcellation method, notably through comparisons to cytoarchitectonic and fMRI data.” [p.69]; “We evaluate how well the parcellations agree with the underlying brain structure by comparing to task fMRI activation maps.” [p.74]; [p.68-72], [fig. 1-4, 8, 11-13], [see claim 1 rejection]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 9, Huo and Parisot teach the method of claim 1,
Huo further teaching wherein the classification of the supervoxels comprises generating weights for the supervoxels using machine learning (“The likelihood score L(s, l) and weight w(s ,l) in our model jointly serve as the label prior and the intensity likelihood, which can be interpreted as the cost of assigning a label l to the supervoxel s from two perspectives. […] Unlike the likelihood score computed from a limited number of local candidates, the weight w(s, l) measures the distance from s to the center of class l by involving the samples randomly selected from all of the supervoxels in the atlases.” [p.205]; “Since each type of feature shows different importance in differentiating tissues, we assign a weight w to each type of feature to improve the accuracy of the supervoxel matching.” [p.206]; [p.203-209], [fig. 1-4, 6, 8, 10-11]).
Regarding claim 10, Huo and Parisot teach the method of claim 9,
Huo further teaching wherein the machine learning comprises gradient boosting decision trees, artificial neural networks, or support vector machines (“The SVM classifier is used to generate the predicted label image of the target for supervoxel segmentation and the probability map for initialization of data term in dense labeling.” [p.203, fig.1 inset]; “In the implementation of the supervoxel segmentation, voxel feature vectors randomly selected from the unregistered atlas intensity images are used as the training samples to train the SVM classifier while the voxel feature vectors extracted from each voxel in the target intensity image are treated as the testing samples.” [p.205]; [p.203-209], [fig. 1-4, 6, 8, 10-11]).
Regarding claim 11, Huo and Parisot teach the method of claim 9,
Parisot further teaching wherein the anatomical location of the functional task is determined based upon the generated weights (“Inter-subject edges are created between the matched supervertices VLSi and VLopt , Sj and weighted as the correlation between the low dimensional merged connectivity profiles associated with each supervertex: […] Weighting the edges with the correlation between the matched supervertices allows control of how similar two parcellations are expected to be locally, based on the similarity of the two subjects' underlying data.” [p.73]; [p.77-83], [fig. 1-4, 8, 11-13]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 12, Huo teaches a system (“The paper proposes a new segmentation framework based on supervoxels” [abst]; “multi-atlas segmentation describes the parcellation procedure through a probabilistic model constructed on a limited number of manually delineated atlas images” [p.201]; [p.202-206], [fig. 1-4, 6, 8, 10-11]), comprising:
at least one computing device comprising processing circuitry including a processor and memory (“The average training time for the SVM classifier is 2 h on a single desktop PC (i7-4900 CPU 3.60 GHz, 16 GB RAM).” [p.212]), the at least one computing device configured to at least:
preprocess data obtained from brains of multiple subjects to generate one or more dynamic parcellated supervoxel maps based on a whole brain map of the multiple subjects (“Supervoxel segmentation is the first step in the supervoxel graph construction” [p.203]; “Moreover, to calculate the data term, each supervoxel in the atlases should have the unique label. Therefore, a refinement scheme is proposed following the supervoxel segmentation of the atlases.” [p.204]; “The proposed approach is thoroughly evaluated on three publically available brain datasets for the whole brain segmentation” [p.206]; “Before applying our approach, we performed the pre-processing steps in the following order for all the tests: bias field correction, pairwise registration, and image normalization. […] In the multi-atlas segmentation, both the intensity images and the label images of the atlases are required to be warped to the target domain with the same transformation. […] 1) The standard scale landmarks corresponding to each decile were calculated using the warped atlas intensity images; and 2) the image histogram (target and atlas intensity images) was mapped to the standard scale landmarks in a piece-wise linear fashion.” [p.207]; [p.203-209], [fig. 1-4, 6, 8, 10-11]); and
determine an anatomical location in a brain of another subject based upon classification of supervoxels of the one or more dynamic parcellated supervoxel maps (“The SVM classifier is used to generate the predicted label image of the target for supervoxel segmentation and the probability map for initialization of data term in dense labeling.” [p.203, fig. 1 inset]; “In the implementation of the supervoxel segmentation, voxel feature vectors randomly selected from the unregistered atlas intensity images are used as the training samples to train the SVM classifier while the voxel feature vectors extracted from each voxel in the target intensity image are treated as the testing samples. […] By applying the SVM classifier to the testing samples, a predicted label image and a probability map of the target are generated simultaneously.” [p.205]; “Unlike the other baseline methods, the result of the proposed method does not contain the undesired “holes” in the anatomical structures, which indicates the advantages of our approach in maintaining the spatial consistency […] We randomly selected 9 subjects as the atlases and the remaining as the test subjects. The SVM classifier with c ¼ 22 and γ ¼ 20 is used to obtain the predicted label image” [p.211]; “The observed qualitative performance improvement demonstrates that the introduction of supervoxels contributes to preserving the spatial label consistency. As shown in the MICCAI 2012 dataset with 134 anatomical structures, the proposed method highlights its benefits in maintaining the spatial label consistency, particularly for the small structures.” [p.212]; The SVM classifier is trained to perform segmentation on multiple atlases and applied to individual training subjects to test the efficacy of the trained model [p.205-213], [fig. 1-4, 6, 8, 10-11]);
but Huo may fail to teach activation data associated with a functional task.
However, in the same field of endeavor, Parisot teaches a system configured to preprocess activation data obtained from brains of multiple subjects to generate one or more dynamic parcellated supervoxel maps based on a whole brain map of the multiple subjects, the activation data associated with a functional task (“The structural and diffusion data have been preprocessed following the HCP's minimal preprocessing pipelines” [p.70]; “The task fMRI data is preprocessed following the HCP preprocessing pipelines (gradient unwarping, motion and distortion correction, registration to the MNI space and projection to the cortical surface). Task activation maps are then obtained using standard FSL tools (FEAT) that use general linear modelling to construct activation maps. […] The analysis is carried out across sessions (single subject activation maps) and then across subjects (group-wise activation map)” [p.76]; [p.69-72], [fig. 1-4, 8, 11-13]); and
determine an anatomical location of the functional task in a brain of another subject based upon classification of supervoxels of the one or more dynamic parcellated supervoxel maps (“we observe strong correspondences between our parcels' boundaries and strong variations of myelination, specifically in the motor areas, both on the average map and the single subject level […] On average over all parcellation resolutions, we obtain very good overlap measurements with the motor areas (BA 1–6),” [p.77]; “At the single subject level, this allows to estimate which brain regions are the most consistent (inter-subject variability), while the group level enables to evaluate the fundamental differences in connectivity and function between two different groups.” [p.81]; [p.77-83], [fig. 1-4, 8, 11-13], [see claim 1 rejection]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The state-of-the-art techniques in multi-atlas segmentation, including both patch-based and learning-based methods, are strongly dependent on the pairwise registration or exhibit huge spatial inconsistency (Huo [abst.]). One of the main advantages of using a group-wise parcellation method is the possibility to perform direct comparisons between subjects as well as groups (gender, age or diseased base groups). At the single subject level, this allows to estimate which brain regions are the most consistent (inter-subject variability), while the group level enables to evaluate the fundamental differences in connectivity and function between two different groups (Parisot [p.81]). The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 13, Huo and Parisot teach the system of claim 12,
Huo further teaching wherein preprocessing the activation data comprises:
registering and averaging the activation data of the multiple subjects to produce the whole brain map (“Before implementing the proposed method, pairwise registrations are performed between the target and each atlas. Then, in order to construct the supervoxel graph, supervoxel segmentations are applied on the target and atlases, respectively” [p.203]; [p.205-213], [fig. 1-4, 6, 8, 10-11]); and
generating the one or more dynamic parcellated supervoxel maps from the whole brain map (“a dense labeling step is proposed to acquire the refined label map of the target” [p.203]; “to construct the matching pairs, we randomly select equal supervoxel sample pairs per class from the registered atlases. Then the equal number of non-matching pairs are randomly selected” [p.206]; [p.205-213], [fig. 1-4, 6, 8, 10-11]).
Regarding claim 16, Huo and Parisot teach the system of claim 12,
Parisot further teaching wherein the activation data is acquired through magnetic resonance imaging of the subject while carrying out the functional task (“We present an extended experimental evaluation of the parcellation method, notably through comparisons to cytoarchitectonic and fMRI data.” [p.69]; “We evaluate how well the parcellations agree with the underlying brain structure by comparing to task fMRI activation maps.” [p.74]; [p.68-72], [fig. 1-4, 8, 11-13], [see claim 1 rejection]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 17, Huo and Parisot teach the system of claim 16,
Parisot further teaching wherein the functional task is associated with a foot, a hand, or a mouth of the subject (“For all resolutions, we observe strong correspondences between our parcels' boundaries and strong variations of myelination, specifically in the motor areas, both on the average map and the single subject level.” [p.77]; [p.77-83], [fig. 1-4, 8, 11-13], [see claim 7 rejection]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Regarding claim 18, Huo and Parisot teach the system of claim 12,
Huo further teaching wherein the classification of the supervoxels comprises generating weights for the supervoxels using machine learning (“The likelihood score L(s, l) and weight w(s ,l) in our model jointly serve as the label prior and the intensity likelihood, which can be interpreted as the cost of assigning a label l to the supervoxel s from two perspectives. […] Unlike the likelihood score computed from a limited number of local candidates, the weight w(s, l) measures the distance from s to the center of class l by involving the samples randomly selected from all of the supervoxels in the atlases.” [p.205]; “Since each type of feature shows different importance in differentiating tissues, we assign a weight w to each type of feature to improve the accuracy of the supervoxel matching.” [p.206]; [p.203-209], [fig. 1-4, 6, 8, 10-11]).
Regarding claim 19, Huo and Parisot teach the system of claim 18,
wherein the machine learning comprises gradient boosting decision trees, artificial neural networks, or support vector machines (“The SVM classifier is used to generate the predicted label image of the target for supervoxel segmentation and the probability map for initialization of data term in dense labeling.” [p.203, fig.1 inset]; “In the implementation of the supervoxel segmentation, voxel feature vectors randomly selected from the unregistered atlas intensity images are used as the training samples to train the SVM classifier while the voxel feature vectors extracted from each voxel in the target intensity image are treated as the testing samples.” [p.205]; [p.203-209], [fig. 1-4, 6, 8, 10-11]).
Regarding claim 20, Huo and Parisot teach the system of claim 19,
Parisot further teaching wherein the anatomical location of the functional task is determined based upon the generated weights (“Inter-subject edges are created between the matched supervertices VLSi and VLopt , Sj and weighted as the correlation between the low dimensional merged connectivity profiles associated with each supervertex: […] Weighting the edges with the correlation between the matched supervertices allows control of how similar two parcellations are expected to be locally, based on the similarity of the two subjects' underlying data.” [p.73]; [p.77-83], [fig. 1-4, 8, 11-13]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method for dynamic supervoxel parcellation taught by Huo with the method using the activation data associated with a functional task as taught by Parisot. The proposed method overcomes challenges existing in previous multi-atlas segmentation in terms of the computational efficiency and the dependency on the complicated deformable pairwise registration. The goals are accomplished by utilizing the graphical model associated with the supervoxels to solve the MAP estimation problem defined in multi-atlas segmentation (Huo [p.212]). The proposed method simultaneously estimates subject specific parcellations that have direct correspondences across subjects. Quantitative and qualitative experiments show that group consistency does not reduce the quality of the parcellation on the subject level (Parisot [p.79]).
Claim(s) 5-6 and 14-15 is/are rejected under 35 U.S.C. 103 as being obvious over Huo and Parisot as applied to claims 2 and 13 above, in further view of deCharms (US20130245424A1, 2013-09-19; hereinafter “deCharms”) as provided by Applicant.
Regarding claim 5, Huo and Parisot teach the method of claim 2,
Parisot further teaching comprising mapping the one or more dynamic parcellated supervoxel maps (“single subject parcellations appear to have similar boundaries with Brodmann's cytoarchitectonic motor areas (BA 1, 2, 3a–b, 4a–b and 6).” [p.77]; [p.77-83], [fig. 1-4, 8, 11-13]);
but the combination of references above may fail to explicitly teach masking using a conjunction map.
However, in the same field of endeavor, deCharms teaches a method for functional task prediction with dynamic supervoxel parcellation (“A computer assisted method is provided for diagnosing a condition of a subject wherein said condition is associated with an activation in one or more regions of interest,” [clm 1]; “Following the forward or reverse correlation process, detail predictions may be made of the representational content inherent in the activation of a single voxel, or complex patterns of activation encompassing multiple voxels. This takes place by taking the joint predicted (or decoded) cognitive process associated with the activation level at each voxel, and combining them to produce an overall estimate of cognitive processing taking place at a given moment.” [0519]; [0189-0381], [fig. 1-2]);
deCharms further teaching comprising masking the one or more dynamic parcellated supervoxel maps using a conjunction map (“Another type of pre-processing that may be performed on the input image/volume data may be the selection of voxels corresponding to the brain. […] This process may also include the masking on of voxels determined to be inside the region corresponding to the brain.” [0402]; “a volume mask is generated corresponding to every point in the subject's brain volume that corresponds to a point from the anatomical structure(s) selected by the device user. This volume mask can be overlayed upon the subject's brain images to allow the user to more easily and accurately select the location of a region of interest, or the volume mask can be used as a region of interest itself.” [0531]; [0382-05210], [fig. 1-2]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method taught by Huo and Parisot with masking using a conjunction map as taught by deCharms. A large number of psychiatric, neurological, and neurodegenerative pathologies involve changes of mental states or conditions based upon changes in neurotransmitter and receptor balances. Detection of such changes may allow for diagnosis well ahead of manifestation of severe clinical symptoms, and knowledge of the nature and the extent of such changes is of paramount importance for the determination of therapy (deCharms [0003]). The combined method may provide measurement in real time of fluctuations of physiological activity due to instructions or other stimulation, comparison of these measurements between people or groups, and use of this process in diagnosis (deCharms [0011]).
Regarding claim 6, Huo, Parisot and deCharms teach the method of claim 5,
de Charms further teaching wherein the one or more dynamic parcellated supervoxel maps comprises average beta coefficients (“Percent correct measures may be made in the same fashion for motor or cognitive tasks. These allow the computation of psychophysical parameters such as d′ and beta according to standard methods familiar to one skilled in the art.” [0362]; [0189-0381], [fig. 1-2]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the method taught by Huo and Parisot with masking using a conjunction map as taught by deCharms. A large number of psychiatric, neurological, and neurodegenerative pathologies involve changes of mental states or conditions based upon changes in neurotransmitter and receptor balances. Detection of such changes may allow for diagnosis well ahead of manifestation of severe clinical symptoms, and knowledge of the nature and the extent of such changes is of paramount importance for the determination of therapy (deCharms [0003]). The combined method may provide measurement in real time of fluctuations of physiological activity due to instructions or other stimulation, comparison of these measurements between people or groups, and use of this process in diagnosis (deCharms [0011]).
Regarding claim 14, Huo and Parisot teach the system of claim 13,
Parisot further teaching comprising mapping the one or more dynamic parcellated supervoxel maps (“single subject parcellations appear to have similar boundaries with Brodmann's cytoarchitectonic motor areas (BA 1, 2, 3a–b, 4a–b and 6).” [p.77]; [p.77-83], [fig. 1-4, 8, 11-13]);
but the combination of references above may fail to explicitly teach masking using a conjunction map.
However, in the same field of endeavor, deCharms teaches a system (“Computer executable software, the software comprising: logic for taking activity measurements of one or more localized brain regions […] based on measured brain activity in substantially real time relative to when the intervention is performed,” [clm 25]; “Following the forward or reverse correlation process, detail predictions may be made of the representational content inherent in the activation of a single voxel, or complex patterns of activation encompassing multiple voxels. This takes place by taking the joint predicted (or decoded) cognitive process associated with the activation level at each voxel, and combining them to produce an overall estimate of cognitive processing taking place at a given moment.” [0519]; [0189-0381], [fig. 1-2], [see claim 5 rejection]);
deCharms further teaching wherein preprocessing further comprises masking the one or more dynamic parcellated supervoxel maps using a conjunction map (“Another type of pre-processing that may be performed on the input image/volume data may be the selection of voxels corresponding to the brain. […] This process may also include the masking on of voxels determined to be inside the region corresponding to the brain.” [0402]; “a volume mask is generated corresponding to every point in the subject's brain volume that corresponds to a point from the anatomical structure(s) selected by the device user. This volume mask can be overlayed upon the subject's brain images to allow the user to more easily and accurately select the location of a region of interest, or the volume mask can be used as a region of interest itself.” [0531]; [0382-05210], [fig. 1-2]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the system taught by Huo and Parisot with masking using a conjunction map as taught by deCharms. A large number of psychiatric, neurological, and neurodegenerative pathologies involve changes of mental states or conditions based upon changes in neurotransmitter and receptor balances. Detection of such changes may allow for diagnosis well ahead of manifestation of severe clinical symptoms, and knowledge of the nature and the extent of such changes is of paramount importance for the determination of therapy (deCharms [0003]). The combined method may provide measurement in real time of fluctuations of physiological activity due to instructions or other stimulation, comparison of these measurements between people or groups, and use of this process in diagnosis (deCharms [0011]).
Regarding claim 15, Huo, Parisot and deCharms teach the system of claim 14,
de Charms further teaching wherein the one or more dynamic parcellated supervoxel maps comprises average beta coefficients (“Percent correct measures may be made in the same fashion for motor or cognitive tasks. These allow the computation of psychophysical parameters such as d′ and beta according to standard methods familiar to one skilled in the art.” [0362]; [0189-0381], [fig. 1-2]).
It would have been obvious to one of ordinary skill in the art prior to the effective filing date of the invention to combine the system taught by Huo and Parisot with masking using a conjunction map as taught by deCharms. A large number of psychiatric, neurological, and neurodegenerative pathologies involve changes of mental states or conditions based upon changes in neurotransmitter and receptor balances. Detection of such changes may allow for diagnosis well ahead of manifestation of severe clinical symptoms, and knowledge of the nature and the extent of such changes is of paramount importance for the determination of therapy (deCharms [0003]). The combined method may provide measurement in real time of fluctuations of physiological activity due to instructions or other stimulation, comparison of these measurements between people or groups, and use of this process in diagnosis (deCharms [0011]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Swisher et al. (WO2020033566A1, 2020-02-13) teaches systems and methods for predicting and managing risk of brain function, disorders, and diseases for individuals [0002].
Soltaninejad et al. (Supervised learning based multimodal MRI brain tumour segmentation using texture features from supervoxels, Computer Methods and Programs in Biomedicine, v157, 2018, p.69-84, ISSN 0169-2607, 2018-02-21) teaches two innovative mechanisms. First, two auxiliary transformations are elaborated that explicitly capture intermodal dependencies: the shallow transformation consolidates spatial intricacies, while the high-dimensional counterpart delves into deep feature extraction through enriched channel representations. Second, based on a meticulous analysis of the non-uniform frequency distribution in the low- and high-frequency components of the mask, we design a new unrolling paradigm that leverages a progressive masking scheme, integrating a dilation–contraction mechanism to dynamically regulate learnable frequency regions during the forward propagation across successive stages [abst].
Hu et al. (US20180286041A1, 2018-10-04) teaches novel methods of accurately and efficiently reconstructing parameter maps in MRI data [abst].
del Re et al. (A New MRI Masking Technique Based on Multi-Atlas Brain Segmentation in Controls and Schizophrenia: A Rapid and Viable Alternative to Manual Masking. J Neuroimaging. 2016 Jan-Feb;26(1):28-36. doi: 10.1111/jon.12313, 2015-11-20) teaches a new brain masking technique based on multi-atlas brain segmentation (MABS) [abst].
Any inquiry concerning this communication or earlier communications from the examiner should be directed to James F. McDonald III whose telephone number is (571)272-7296. The examiner can normally be reached M-F; 8AM-6PM 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, Chris Koharski can be reached at 5712727230. 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.
JAMES FRANKLIN MCDONALD III
Examiner
Art Unit 3797
/CHRISTOPHER KOHARSKI/Supervisory Patent Examiner, Art Unit 3797
Prosecution Insights