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 .
Amendments of record are entered.
Claims 1-37 are pending, of which claims 7-35 are withdrawn.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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.
Claim(s) 1-6, 36-37 is/are rejected under 35 U.S.C. 103 as being unpatentable over Tegzes et al. (US 2018/0315188) in view of Wang (US 2022/0366259) in view of Price (2018/0108137).
As to claim 1:
Tegzes discloses:
One or more processors comprising:
circuitry one or more circuits to: use one or more neural networks to segment one or more objects of one or more images (¶0125, 0055, 0124, 0134, neural networks to perform the method and driven by processors/controller circuitry), wherein at least one of the one or more segmentations of the one or more neural networks are generated for the one or more objects associated with one or more annotated bounding boxes of the one or more images. (See Abstract, ¶0114, determining/generating bounding boxes, each capturing a suspected object in a respective region of interest. ¶0115, once bounding box is determined, the image data in the bounding box is segmented. See also Fig. 19, step 1904 through 1910. See ¶0059, 0065, 0108, 0121 one or more neural networks used for the object processing of the system, i.e. segmentation and classifications. In particular, ¶0108, forming bound box step is performed using a trained deep CNN. ¶0145, ground truth box, ¶0181, reference region is/are labelled for comparison/reference),
Tegzes is, however, silent on “the one or more neural networks comprising a student neural network that comprises parameters that are updated based, at least in part, on a comparison of one or more segmentations of a teacher neural network with one or more segmentations of the student neural network to evaluate segmentation accuracy of the student neural network as compared to the teacher neural network”
The limitation above is directed to a known methodology of training/retraining of a neural network using knowledge distillation (i.e. teacher supervising student’s performance to further refine the student network), which is disclosed in at least reference Wang as below:
Wang, in a related of field of image segmentation (¶0059), discloses a machine learning model having a teacher network and a student network performing segmentation processing on the input image(s) to produce respective output (for example per ¶0100, 0059), wherein the student network is modified/updated to ensure its output to be consistent with the teacher network via a consistency check (i.e. a comparison process), and also further based on the loss function value from the teacher network (¶0106, 0065, also 0091-0093 discussing the supervising by the teacher network’s output to improve the student network’s parameter).
It would have been obvious to one of ordinary skill in the art before the effective filing time of the invention that the operational neural network performing segmentation in Tegzes is implemented to be trained/updated by a teacher network in similar manner disclosed by Wang. This implementation advantageously benefits and improve performance of the student networks per ¶0015 of Wang, (“ the student neural network understands the training process of the teacher neural network more fully, thereby effectively improving the performance (e.g. accuracy) of the student network”)
Tegzes and Wang do not explicitly mention bounding boxes and one or more annotations are comprised of inputs for the generation of the at least one of the one or more segmentation of the teacher neural network and the one or more segmentation of the student neural network.
Price, however in a related field of semantic segmentation in supervised manner, discloses an image segmentation process/system (Abstract) wherein, the segmentation model is comprised of at least one neural network to perform segmentation, wherein inputs for the model include a set of bounding boxes established in the image input and annotation (i.e. object information including class labels corresponding to the target objects corresponding to the bounding boxes, where the class level annotates the class of the image objects) (See at least ¶0050-0053, 0021-0023).
It would have been obvious to one of ordinary skill in the art before the effective filing time of the invention that the segmentation performed by either teacher/student network in the system of Tegzes in view of Wang wherein mention bounding boxes and one or more annotations are comprised of inputs for the generation of the at least one of the one or more segmentation of the teacher neural network and the one or more segmentation of the student neural network. Tegzes in ¶0050, 0051 explicitly states the neural networks are trained and operate in supervised manner, i.e. with label-based guidance, thus deeply connected to disclosure of Price, also a supervised training. This implication of image analysis with annotation (i.e. labels), i.e. supervised process, can provide sufficient accuracy and performance goal and reduce error (i.e. false classification) per Tegzes, ¶0069, Price, ¶0058)
As to claim 2:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein the one or more segmentations of the student network comprises generating one or more segmentation maps indicating pixels of the one or more objects. (See Tegzes, ¶01030, 0133, ¶0177, output comprising generation of output with each pixel/voxel identified/labelled)
As to claim 3:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein the circuitry is to use at least one of the teacher neural network and the student neural network to determine one or more classifications of the one or more objects. (See Tegzes, ¶0100, 0108, 0116, 0188, using the CNNs to classify the object(s) as a particular organ or an organ of interest)
As to claim 4:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein the circuitry is to use at least one of the teacher neural network and the student neural network to determine instances of the one or more objects in the one or more images. (Tegzes, ¶0104, the system configured to detect presence of objects (instances) to determine whether they are relevant. ¶0137, determine more than one object of interests for identification)
As to claim 5:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein the circuitry is to use at least one of the teacher neural network and the student neural network to determine correspondence between the one or more objects in the one or more images. (Tegzes, ¶0140, 0155, for example, for paired organs (eyeballs) two similar size boxes, as the system determines they are separate instances but related as the same type of organ. ¶0159, determining whether objects in images are related, for example objects of the body are determined as the set of object to be considered, whereas object outside the body are determined as irrelevant)
As to claim 6:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein the circuitry is cause the teacher neural network to generate one or more labels associated with the one or more images to be used for the one or more segmentations of the student network. (Tegzes, ¶0177, “provide a semantic segmentation (e.g., pixel-wise classification) of image data to label each pixel and segment the image into regions corresponding to sets of labels. Similarly, each voxel can be classified according to a region of interest to which it belongs. An example classifier can process image data on a voxel-by-voxel level to provide an output for each voxel based on its neighboring voxels and classify an entire image slice in a single action”)
As to claim 36:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein at least one of the teacher neural network and the student neural network are to generate one or more proposals for one or more locations of the one or more annotated bounding boxes. (See Tegzes, ¶0112-0114, proposal of several boundaries/parameters of the current proposed bounding boxes are used to determine precisions.)
As to claim 37:
Tegzes in view of Wang and Price discloses all limitations of claim 1, wherein the one or more annotations indicate at least one of a location or a classification of the one or more object. (See Price, ¶0051, annotation indicating class, Tegzes, ¶0182, “ image labelling in which portions of each image 2702-2712 are associated with an appropriate label 2720 identifying the anatomy, organ, region, or object of interest (e.g., brain, head (and neck), chest, upper abdomen, lower abdomen, upper pelvis, center pelvis, lower pelvis, thigh, shin, foot, etc.).”.)
Response to Arguments
Applicant’s arguments with respect to the claim(s) have been considered but are moot because a ground of rejection is established with new reference(s) addressed to new limitations and are not specifically challenged in the argument.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Hassan et al. (US 2021/0375008) - Disclosed in some examples are methods, systems, devices, and machine-readable mediums which encode data into a geometric representation for more efficient and secure processing. For example, data may be converted from a binary representation to a geometric representation using an encoding dictionary. The encoding dictionary specifies one or more geometric shapes used in the encoding. The geometrically encoded data may comprise one or more identifiers that specify one or more of the shapes of the encoding dictionary that best match one or more detected features in an image corresponding to the data. In some examples, the geometrically encoded data may also comprise one or more transformations of the one or more shapes to reduce error in the geometric encoding.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to QUAN M HUA whose telephone number is (571)270-7232. The examiner can normally be reached 10:30-6:30.
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, Anthony Addy can be reached at 571-272-7795. 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.
/QUAN M HUA/Primary Examiner, Art Unit 2645
Prosecution Insights