DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed.
Response to Amendment
Claims 1-20 were previously pending and subject to non-final action filed on 01/13/2026. In the response filed 03/04/2026, claims 1, 19 and 20 were amended. Therefore, claims 1-20 are currently pending and subject to the final action below.
Response to Arguments
Applicant’s arguments, see page 10, filed 03/04/2026, with respect to the Drawing have been fully considered and are persuasive. The objection of the Drawing has been withdrawn.
Applicant’s arguments, see page 10, filed 03/04/2026, with respect to claims 20 under 35 U.S.C. 101 rejection have been fully considered and are persuasive. The 101 rejection of the claim 20 has been withdrawn.
Applicant's arguments, see page 11-17, filed 03/04/2026 with respect to claims 1-20 under 35 U.S.C. 103 have been fully considered but they are not persuasive.
Applicant’s argument: With respect to claims 1, 19 and 20, Applicants respectfully submit that the asserted combination of Georgescu in view of MICHIELS fails to teach or suggest at least: providing a sparse sampling model for sparse sampling the volumetric medical image, wherein the sparse sampling model defines a number of non-adjacent sampling points distributed in the volumetric medical image and defines locations and distances of the distributed sampling points; sampling non-adjacent voxels from the volumetric medical image using the provided sparse sampling model for obtaining sparse sampling descriptors; classifying labels for query points in the volumetric medical image by applying a trained classifier to the sparse sampling descriptors. First, the asserted combination of Georgescu and MICHIELS fails to teach or suggest at least providing a sparse sampling model for sparse sampling the volumetric medical image, wherein the sparse sampling model defines a number of non-adjacent sampling points distributed in the volumetric medical image and defines locations and distances of the distributed sampling points.
Georgescu only describes generation of image patches that are used as hypotheses in the current search space to train the deep neural network for that search space. See Georgescu at paragraph [0041]. These operations do not constitute sparse sampling. Patch extraction involves contiguous voxel blocks, not non-adjacent sampling points defined by a spatial sampling model. Moreover, such "image patches" are input into the sparse deep neural networks (SADNN) to train them, and not output by the SADNN. See Georgescu at paragraph [0089]. The SADNN are used to segment a 3D anatomical object in an input 3D medical image, not to define a number of non-adjacent sampling points distributed in the volumetric medical image and locations and distances of the distributed non-adjacent sampling points. Second, the asserted combination of Georgescu and MICHIELS fails to teach or suggest at least sampling non-adjacent voxels from the volumetric medical image using the provided sparse sampling model for obtaining sparse sampling descriptors. Third, the asserted combination of Georgescu and MICHIELS fails to teach or suggest at least classifying labels for query points in the volumetric medical image by applying a trained classifier to the sparse sampling descriptors.
Finally, Applicants respectfully submit that the rejection improperly uses Applicants' claims as a roadmap to reconstruct the claims from disparate documents. In determining obviousness, both pre-AIA and AIA § 103 expressly require considering the claimed invention "as a whole." The question under 35 U.S.C. 103 is not whether the differences themselves would have been obvious, but whether the claimed invention as a whole would have been obvious. See MPEP §2141.02 (emphasis added).3
Examiner Response: After careful consideration the examiner respectfully disagrees.
Applicant’s argument improperly imports limitations from selected embodiments of the specification into the claims. The claims recite “non-adjacent sampling points” and “non-adjacent voxels” but do not require a particular grid arrangement, nested grid structure, spacing value, sampling frequency, or every-second-voxel sampling pattern. The claims further do not require that the sampling points be generated according to any specific geometric model. Under the broadest reasonable interpretation of “non-adjacent” encompasses sampling locations that are spatially separated from one another within the volumetric image.
Georgescu teaches: a) receiving a volumetric medical image comprising at least one organ or portion thereof; (Georgescu − [0039] Referring to FIG. 1, at step 102, training images are received. In particular, a plurality of training images are loaded from a database. The training images can be 2D or 3D medical images acquired using any medical imaging modality, such as but not limited to CT, MRI, Ultrasound, X-ray fluoroscopy, DynaCT, etc. At least a subset of the training images are annotated with the pose (e.g., position, orientation, and scale) of the target anatomical object. The training images may also include non-annotated images as well.) Examiner Note: the 3D medical images are a volumetric medical image.
b) providing a sparse sampling model for sparse sampling the volumetric medical image, (Georgescu – Fig. 4, 5 and 21, 26 [0057] At step 412, the position-orientation hypotheses are passed through the trained second deep neural network and a number of best position-orientation candidates are kept. For the discriminative deep neural network, a number of position-orientation hypotheses having the highest probability as calculated by the trained second deep neural network can be kept as the position-orientation candidates for each training image. [0062] At step 506, the position candidates detected by the first trained deep neural network are augmented with orientation parameters to generate position-orientation hypotheses. For example, a plurality of position-orientation hypotheses can be generated for each detected position candidate by rotating each image patch centered at a position candidate to a plurality of possible orientations sampled from a predetermined range of orientations for the target anatomical object. anatomical object can be displayed on a display device of a computer. FIG. 24 illustrates exemplary results for detecting the aortic valve in 3D ultrasound images using the method of FIG. 21. [0119] At step 2604, landmark candidates for a target anatomical landmark are detected in the 3D medical image using an initial shallow neural network landmark detector.)
Georgescu teaches evaluating a plurality of position candidates, position-orientation hypotheses, landmark candidates, and boundary points distributed throughout a volumetric medical image. These locations are spatially separated by intervening image locations. Therefore corresponds to the claimed “non-adjacent sampling points under the BIR of the claim limitations.
wherein the sparse sampling model defines a number of non-adjacent sampling points distributed in the volumetric medical image and defines locations and distances of the distributed non-adjacent sampling points; (Georgescu − [0112] A mean model (e.g., 3D mesh) of the target anatomical object can be calculated from a set of annotated training data, and once the full parameter set (position, orientation, and scale) of the target anatomical object is detected in the medical image. Fig. 21 The trained SADNN boundary detector calculates a probability for each point sampled along the normal line, and the boundary is refined by moving each boundary point to the point along the normal line having the highest probability. [0114] The segmented 3D mesh of the target anatomical object and/or the bounding box detection result for the target anatomical object can be displayed on a display device of a computer. FIG. 24 illustrates exemplary results for detecting the aortic valve in 3D ultrasound images using the method of FIG. 21.)
Georgescu teaches: evaluating a plurality of spatially distributed candidate locations and boundary points within a volumetric medial image. These points are separated by intervening locations and therefore constitute non-adjacent sampling points under a BRI of the claim limitations.
c) sampling non-adjacent voxels from the volumetric medical image using the provided sparse sampling model for obtaining sparse sampling descriptors; (Georgescu − [0121] In object detection using a sliding window based approach, for each position hypothesis, an image patch (with a pre-defined size) centered at the position hypothesis is cropped. The patch intensities are then serialized into a vector to calculate a response. Repeat the process by shift the image patch by one voxel.)
Georgescu teaches: for each position hypothesis and candidate location, an extraction and processing of image data associated with those selected position hypotheses within the volumetric image. Only selected candidate locations are evaluated rater than every voxel location in the image volume. Therefore, Georgescu teaches sampling a subset of voxels associated with spatially distributed non-adjacent candidate locations.
MICHIELS teaches: d) classifying labels for query points in the volumetric medical image by applying a trained classifier to the sparse sampling descriptors; (MICHIELS − [0058] Segmentation is task of assigning a specific label of each part of the input to identify an anatomical structure and/or anatomical landmarks. [0078] Anatomical landmark identification module 114 then may automatically detect the ostium plane centroid (OPC) using the point detection deep learning module. For example, the landmark detection deep learning model may be any deep learning architecture such as Dense V-Net or SegResNet. [0096] For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module.) assign label to anatomical landmarks in 3D reconstruction image.
and e) providing a segmentation mask for the volumetric medical image using the classified labels. (MICHIELS − [0011] generating a segmentation mask comprising each voxel of the plurality of voxels assigned the predetermined label. [0096] For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module.)
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Examiner Note
The term “sparse sampling/points is understood as reduced number of data points or measurements. The term “query point” are location of known and unknown signals, function, or value. The term “segmentation mask” are also called pixel-wise, pixel-wise maps, pixel-wise class maps, segmentation maps, binary masks, object masks, image patch or simply masks. Voxel is essentially a 3D pixel, therefore a pixel in 3D can be interpreted as a voxel.
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.
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) 1-6 and 8-20 are rejected under 35 U.S.C. 103 as being unpatentable over Georgescu (US 20160174902 A1, Filed Date: Feb. 26, 2016) in view of MICHIELS (US 20230119535 A1, Filed Date: Oct. 12, 2022).
Regarding independent claim 1, Georgescu teaches: A computer-implemented method for automated processing of volumetric medical images, the method comprising:
: a) receiving a volumetric medical image comprising at least one organ or portion thereof; (Georgescu − [0039] Referring to FIG. 1, at step 102, training images are received. In particular, a plurality of training images are loaded from a database. The training images can be 2D or 3D medical images acquired using any medical imaging modality, such as but not limited to CT, MRI, Ultrasound, X-ray fluoroscopy, DynaCT, etc. At least a subset of the training images are annotated with the pose (e.g., position, orientation, and scale) of the target anatomical object. The training images may also include non-annotated images as well.) Examiner Note: the 3D medical images are a volumetric medical image.
b) providing a sparse sampling model for sparse sampling the volumetric medical image, (Georgescu – Fig. 4, 5 and 21, 26 [0057] At step 412, the position-orientation hypotheses are passed through the trained second deep neural network and a number of best position-orientation candidates are kept. For the discriminative deep neural network, a number of position-orientation hypotheses having the highest probability as calculated by the trained second deep neural network can be kept as the position-orientation candidates for each training image. [0062] At step 506, the position candidates detected by the first trained deep neural network are augmented with orientation parameters to generate position-orientation hypotheses. For example, a plurality of position-orientation hypotheses can be generated for each detected position candidate by rotating each image patch centered at a position candidate to a plurality of possible orientations sampled from a predetermined range of orientations for the target anatomical object. anatomical object can be displayed on a display device of a computer. FIG. 24 illustrates exemplary results for detecting the aortic valve in 3D ultrasound images using the method of FIG. 21. [0119] At step 2604, landmark candidates for a target anatomical landmark are detected in the 3D medical image using an initial shallow neural network landmark detector.)
Georgescu teaches evaluating a plurality of position candidates, position-orientation hypotheses, landmark candidates, and boundary points distributed throughout a volumetric medical image. These locations are spatially separated by intervening image locations. Therefore corresponds to the claimed “non-adjacent sampling points under the BIR of the claim limitations.
wherein the sparse sampling model defines a number of non-adjacent sampling points distributed in the volumetric medical image and defines locations and distances of the distributed non-adjacent sampling points; (Georgescu − [0112] A mean model (e.g., 3D mesh) of the target anatomical object can be calculated from a set of annotated training data, and once the full parameter set (position, orientation, and scale) of the target anatomical object is detected in the medical image. Fig. 21 The trained SADNN boundary detector calculates a probability for each point sampled along the normal line, and the boundary is refined by moving each boundary point to the point along the normal line having the highest probability. [0114] The segmented 3D mesh of the target anatomical object and/or the bounding box detection result for the target anatomical object can be displayed on a display device of a computer. FIG. 24 illustrates exemplary results for detecting the aortic valve in 3D ultrasound images using the method of FIG. 21.)
Georgescu teaches: evaluating a plurality of spatially distributed candidate locations and boundary points within a volumetric medial image. These points are separated by intervening locations and therefore constitute non-adjacent sampling points under a BRI of the claim limitations.
c) sampling non-adjacent voxels from the volumetric medical image using the provided sparse sampling model for obtaining sparse sampling descriptors; (Georgescu − [0121] In object detection using a sliding window based approach, for each position hypothesis, an image patch (with a pre-defined size) centered at the position hypothesis is cropped. The patch intensities are then serialized into a vector to calculate a response. Repeat the process by shift the image patch by one voxel.)
Georgescu teaches: for each position hypothesis and candidate location, an extraction and processing of image data associated with those selected position hypotheses within the volumetric image. Only selected candidate locations are evaluated rater than every voxel location in the image volume. Therefore, Georgescu teaches sampling a subset of voxels associated with spatially distributed non-adjacent candidate locations.
Georgescu teaches segmentation mask for volumetric medical images (image patch) but does not explicitly teach: classifying labels in the volumetric medical image by applying a trained classifier;
However, MICHIELS teaches: d) classifying labels for query points in the volumetric medical image by applying a trained classifier to the sparse sampling descriptors; (MICHIELS − [0058] Segmentation is task of assigning a specific label of each part of the input to identify an anatomical structure and/or anatomical landmarks. [0078] Anatomical landmark identification module 114 then may automatically detect the ostium plane centroid (OPC) using the point detection deep learning module. For example, the landmark detection deep learning model may be any deep learning architecture such as Dense V-Net or SegResNet. [0096] For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module.) assign label to anatomical landmarks in 3D reconstruction image.
and e) providing a segmentation mask for the volumetric medical image using the classified labels. (MICHIELS − [0011] generating a segmentation mask comprising each voxel of the plurality of voxels assigned the predetermined label. [0096] For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module.)
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 2, depends on claim 1, Georgescu teaches: further comprising identifying type of organ at a point of interest. (Georgescu − [0079] FIG. 9 illustrates a method for left ventricle (LV) landmark detection in MR cardiac long axis image [0135] In an advantageous embodiment, the method of FIG. 26 is applied to automatic carotid artery bifurcation landmark detection in head-neck CT scans. The carotid artery is the main vessel supplying oxygenated blood to the head and neck.)
Regarding dependent claim 3, depends on claim 2, Georgescu teaches: wherein the type of organ is identified by applying the trained classifier to the sampled voxels. (Georgescu − [0079] FIG. 9 illustrates a method for left ventricle (LV) landmark detection in MR cardiac long axis image [0133] Returning to FIG. 26, at step 2608, the landmark is detected from the landmark candidates based on the deeply learned (neural network) features and other features extracted from the 3D medical image using a trained classifier [0135] In an advantageous embodiment, the method of FIG. 26 is applied to automatic carotid artery bifurcation landmark detection in head-neck CT scans. The carotid artery is the main vessel supplying oxygenated blood to the head and neck.)
Regarding dependent claim 4, depends on claim 2, Georgescu does not explicitly teach: further comprising: receiving a command
However, MICHIELS teaches: further comprising: receiving a command for determining the point of interest, wherein the sparse sampling model is provided in dependence on the received command. (MICHIELS − In addition, the computerized method further may include receiving user input feedback based on the displayed virtual three-dimensional model; and adjusting the anatomical measurements based on the user input feedback. [0054] User interface 108 may be used to receive inputs from, and/or provide outputs to, a user. For example, user interface 108 may include a touchscreen, display, switches, dials, lights, etc. Feedback input to an ROI and reprocessing the data through deep learning model.)
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 5, depends on claim 4, Georgescu teaches: wherein the sparse sampling model is provided such that the voxels are sampled with a sampling rate per unit length, area or volume which decreases with a distance of a respective voxel on the point of interest. (Georgescu − [0041] randomly selecting image patches using adjust parameters for image patch size and image patch distance from the ground truth center position of the target anatomical object. Image patch distance and size can change by decreasing distance and image patch size.)
Regarding dependent claim 6, depends on claim 1, Georgescu teaches: wherein the sparse sampling model defines a plurality of grids of different grid spacings, the different grid spacings determining different distances of the distributed sampling points in the volumetric medical image. (Georgescu − [0041] randomly selecting image patches using adjust parameters for image patch size and image patch distance from the ground truth center position of the target anatomical object. Image patch distance and size can change by decreasing distance and image patch size.)
Regarding dependent claim 8, depends on claim 1, Georgescu does not explicitly teach: classifying the labels
However, MICHIELS teaches: further comprising: receiving a query determining the query points for classifying the labels to the sparse sampling descriptors, the query defining the locations and distances of the query points in the volumetric medical image. (MICHIELS − [0058] Segmentation is task of assigning a specific label of each part of the input to identify an anatomical structure and/or anatomical landmarks. [0078] Anatomical landmark identification module 114 then may automatically detect the ostium plane centroid (OPC) using the point detection deep learning module. For example, the landmark detection deep learning model may be any deep learning architecture such as Dense V-Net or SegResNet. [0096] For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module.) assign label to anatomical landmarks in 3D reconstruction image.
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 9, depends on claim 8, Georgescu does not explicitly teach: receiving a command for adjusting the query
However, MICHIELS teaches: receiving a command for adjusting the query, wherein the locations and distances of the query points of the query are adjusted in dependence on the received command. (MICHIELS − [0054] User interface 108 may be used to receive inputs from, and/or provide outputs to, a user. Moreover, user interface 108 may receive user input, as well as feedback from the user based on the displayed information, e.g., corrected measurements, such that platform 100 may adjust the information accordingly.)
Regarding dependent claim 10, depends on claim 1, Georgescu does not explicitly teach: wherein steps d) and e) include: receiving a first query determining first query points for providing a coarse segmentation mask,
However, MICHIELS teaches: wherein steps d) and e) include: receiving a first query determining first query points for providing a coarse segmentation mask, the first query defining first locations and first spacings of the first query points in the volumetric medical image, (MICHIELS − [0072] obtaining a 3D point of voxel at predetermined label; input a smaller coarse grained point detection as input data. Corse grained point detection is coarse segmentation mask.)
classifying first labels for the first query points in the volumetric medical image by applying the trained classifier to the sparse sampling descriptors, providing the coarse segmentation mask for the volumetric medical image using the classified first labels, (MICHIELS − [0072] the point detection deep learning module obtains a 3D point by taking a centroid of all voxels having the predetermined label, e.g., with label one, and identifies the specific region of interest of the anatomical structure (the anatomical landmark), e.g., by cropping the MSCT data around the centroid.)
receiving a second query determining second query points for providing a fine segmentation mask, the second query defining second locations and second spacings of the second query points in the volumetric medical image, wherein the second spacings are different to the first spacings, (MICHIELS − [0078] taking centroid points for fine grained resolution 1.0 mm, higher resolution. )
classifying second labels for the second query points in the volumetric medical image by applying the trained classifier to the sparse sampling descriptors, and providing the fine segmentation mask for the volumetric medical image using the classified second labels. (MICHIELS − [0078-0081] fine grained LAA segmentation is output for anatomical land mark identification.)
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 11, depends on claim 10, Georgescu does not explicitly teach: wherein the second spacings are smaller than the first spacings.
However, MICHIELS teaches: wherein the second spacings are smaller than the first spacings. (MICHIELS − [0078-0081] fine grained LAA segmentation is output for anatomical land mark identification. Fine-grained segmentation is high quality imagery with smaller spacing.)
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 12, depends on claim 10, Georgescu does not explicitly teach: wherein the second query is determined such that second query points are selected from neighbors where two of neighbor first query points have different labels.
However, MICHIELS teaches: wherein the second query is determined such that second query points are selected from neighbors where two of neighbor first query points have different labels. (MICHIELS − [0096] Display generation module 122 may be executed by processor 102 for causing a display, e.g., user interface 108 or an external computing device, to display the 3D model reconstruction of the anatomical structure to facilitate preoperative planning by a physician. For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module 118.) different labels for different anatomical parts with different measurements.
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 13, depends on claim 1, Georgescu teaches wherein at least one of the sparse sampling descriptors is formed as a vector of values of the sampled voxels associated to a sampling point of the distributed sampling points. ; (Georgescu − [0121] In object detection using a sliding window based approach, for each position hypothesis, an image patch (with a pre-defined size) centered at the position hypothesis is cropped. The patch intensities are then serialized into a vector to calculate a response. Repeat the process by shift the image patch by one voxel.) receive intensity value for each voxel; the intensity value is the sparse sampling descriptors calculated by the neural network.
Regarding dependent claim 14, depends on claim 1, Georgescu teaches: wherein the provided segmentation mask or part thereof is displayed on a graphical user interface, (Georgescu − [0079] FIG. 9 illustrates a method for left ventricle (LV) landmark detection in MR cardiac long axis image [0135] In an advantageous embodiment, the method of FIG. 26 is applied to automatic carotid artery bifurcation landmark detection in head-neck CT scans. The carotid artery is the main vessel supplying oxygenated blood to the head and neck.)
Georgescu does not explicitly teach: the segmentation mask represents a different label
However, MICHIELS teaches: wherein the segmentation mask is displayed such that each intensity in the segmentation mask represents a different label. (MICHIELS − [0096] Display generation module 122 may be executed by processor 102 for causing a display, e.g., user interface 108 or an external computing device, to display the 3D model reconstruction of the anatomical structure to facilitate preoperative planning by a physician. For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module 118.) different labels for different anatomical parts with different measurements.
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
medical imagery to ensure a fast and accurate anatomical analysis.
Regarding dependent claim 15, depends on claim 1, Georgescu does not explicitly teach: and to provide the labels as an output
However, MICHIELS teaches: wherein the trained classifier comprises a neural network that is configured to receive the sparse sampling descriptors and to provide the labels as an output. (MICHIELS − [0058] Segmentation is task of assigning a specific label of each part of the input to identify an anatomical structure and/or anatomical landmarks. [0078] Anatomical landmark identification module 114 then may automatically detect the ostium plane centroid (OPC) using the point detection deep learning module. For example, the landmark detection deep learning model may be any deep learning architecture such as Dense V-Net or SegResNet. [0096] For example, display generation module 122 may display the 3D model reconstruction overlaid with the identified anatomical landmarks in a manner easily identifiable by the user, e.g., in a different color format or with designated readable labels, as shown in the upper left denoted by (A) of FIG. 6. Display generation module 122 further may display the 3D model reconstruction alongside one or more of the anatomical measurements derived by anatomical measurement determination module.) assign label to anatomical landmarks in 3D reconstruction image.
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 16, depends on 15, Georgescu teaches: wherein the neural network but does not explicitly teach: residual neural network
However, MICHIELS teaches: comprises a residual neural network. (MICHIELS − [0058] Segmentation is task of assigning a specific label of each part of the input to identify an anatomical structure and/or anatomical landmarks. [0078] For example, the landmark detection deep learning model may be any deep learning architecture such as Dense V-Net or SegResNet.) ResNet stands for Residual Network
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 17, depends on 15, Georgescu does not explicitly teach: wherein each of the labels comprises a vector of estimated probabilities for each organ.
However, MICHIELS teaches: wherein each of the labels comprises a vector of estimated probabilities for each organ. (MICHIELS − [0058-0059] . Upon execution of the trained segmentation deep learning module by anatomical landmark identification module 114, a probability mask is returned, which describes the probability that a certain voxel belongs to the anatomical structure label, e.g., the LAA.)
Accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claim invention to combine the teaching of Georgescu and MICHIELS. Both references are directed to automated analysis of volumetric medical images for identifying anatomical structures and landmarks. One of ordinary skill in the art would have been motivated to incorporate the classification framework of MICHIELS into the anatomical detection framework of MICHIELS into the anatomical structures while maintaining efficient processing of volumetric medical image data. Therefore, the combination merely applies known image classification techniques to known anatomical detection techniques to obtain the predictable results of improved anatomical analysis.
Regarding dependent claim 18, depends on 2, Georgescu teaches wherein the sparse sampling descriptors are decoded into a two-dimensional (2D) image including the point of interest, wherein the 2D image is displayed on a graphical user interface together with a plurality of different three-dimensional (3D) slices of the volumetric medical image, (Fig. 14 landmarks of 3D of area of interest; display 2D of area of interest) the plurality of different 3D slices having different resolutions and different ranges. ([0131] image patches for each of the landmark candidates can be extracted on an image pyramid with multiple resolutions)
Regarding independent claim 19, is directed to a computer readable medium. Claim 19 have similar/same technical features/limitations as Claim 1 and the claim is rejected under the same rationale.
Regarding independent claim 20, is directed to a computer readable medium. Claim 19 have similar/same technical features/limitations as Claim 1 and the claim is rejected under the same rationale.
Claim(s) 7 is rejected under 35 U.S.C. 103 as being unpatentable over Georgescu, MICHIELS as applied to claim 1 above, and further in view of JANG (US PGPUB: US 20180018766 A1).
Regarding dependent claim 7, depends on claim 1, Georgescu does not explicitly teach: three-dimensional grids
However, JANG teaches: wherein the plurality of grids comprise three-dimensional grids. (JANG − [0047] In embodiments, the term “voxel” may represent a value on a regular grid in a 3D space and may represent graphic information of a one end which defines one point of the 3D space. For example, each coordinate in three dimensions may represent a location, a color, and density.)
Accordingly, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have combined the teaching of Georgescu, MICHIELS and JANG are directed to automated analysis of volumetric medical images while JANG teaches representing image data using 3D voxel grids. One of ordinary skill in the art would have been motivated to incorporate JANG’s known 3D grid representation within the volumetric medical image framework of Georgescu and MICHIELS since 3D voxel grids are well known structures for representing volumetric image data. The combination applies a known volumetric image representation technique to known anatomical image analysis technique to obtain the predictable result of processing and analyzing anatomical structures within a 3D image volume.
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 CARL E BARNES JR whose telephone number is (571)270-3395. The examiner can normally be reached Monday-Friday 9am-6pm.
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, Stephen Hong can be reached at (571) 272-4124. 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.
/CARL E BARNES JR/Examiner, Art Unit 2178
/STEPHEN S HONG/Supervisory Patent Examiner, Art Unit 2178
Prosecution Insights