DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
Applicant’s arguments and amendments in the Amendment filed July 15, 2026 (herein “Amendment”), with respect to the objection for claiming improper multiple dependency of claims 12, 18 and 25, and claims depending therefrom have been fully considered and are persuasive. The objection to claims 12, 18 and 25, and claims depending therefrom has been withdrawn.
Applicant’s arguments and amendments in the Amendment with respect to the rejection of claim 1, and therefore claims 2–5 which depend therefrom under 35 U.S.C. 112(b) have been fully considered and are persuasive. The rejection of claim 1, and therefore claims 2–5 which depend therefrom under 35 U.S.C. 112(b) has been withdrawn.
Applicant's arguments and amendments in the Amendment regarding the rejection of claims 1–5 under 35 U.S.C. 103 have been fully considered but they are not persuasive. Specifically, Applicant argues on pages 13–14 of the Amendment that because secondary reference Isaksson, relied upon to provide teachings of “generate the difference parameters without requiring the software-tracked contour or the adjusted contour,” teaches that after training, an image contour pair is input to process and output the Dice coefficients (difference parameters) then Isaksson cannot be found to teach “generate the difference parameters without requiring the software-tracked contour or the adjusted contour.” However, Issakson’s model performs inferencing (after training) without requiring (as input) the software-tracked contour or the adjusted contour in that all that is required as input to be provided is an image. As shown in fig. 1 of Isaksson, only during training are the ground truth segmentations also provided as input. Otherwise, during inferencing the input is an image, and a preprocessing step of the model, a segmentation model receives the image and generates output. Applicant’s own invention, as discussed on page 12 of the originally filed Specification “In the inference pipeline step” operates similarly. Therefore, the broadest reasonable interpretation given to the claimed “the difference model uses the input image … without requiring the software-tracked contour or the adjusted contour” is met by the teaching’s of Issakson’s inferencing system which receives just the input image to find the Dice difference parameters.
Applicant also argues on page 14 of the Amendment that “the model designs of Isaksson and the present application are totally different,” and even though, this conclusion is by way of Applicant’s mischaracterization of Isaksson’s model design (which is trained using not just machine segmentations but also manual segmentations by expert medical professionals), nonetheless, Applicant does not state an element of obviousness that “different model designs” would weigh against. That is, Applicant’s arguments do not tie their contention about differences in design between a secondary cited art reference versus that of a disclosed (not necessarily claimed) invention, and how that would effect the prima facie case of obviousness. Therefore, these arguments are not persuasive for not being relevant to the considerations of obviousness.
Accordingly, in view of the above, while all of Applicant’s arguments regarding the rejection of claims 1–5 have been fully considered, they are not persuasive and the rejection is herein maintained.
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.
Claims 1–5 are rejected under 35 U.S.C. 103 as being unpatentable over Sirjani et al., “Automatic cardiac evaluations using a deep video object segmentation network,” Insights Imaging 13, 69 (April 8, 2022), https://doi.org/10.1186/s13244-022-01212-9 (herein “Sirjani”) in view of Isaksson et al., “Quality assurance for automatically generated contours with additional deep learning,” Insights into Imaging, 2022, 13:137, https://doi.org/10.1186/s13244-022-01276-7 (herein “Isaksson”).
Regarding claim 1, with deficiencies noted in square brackets [], Sirjani teaches a method of training a difference model (Sirjani pages 3–4, figs. 1 and 2, training of the EchoRCNN (model)) to generate difference parameters related to the differences between a software-tracked contour of an object and an adjusted contour of the object in an input image or an input video (Sirjani pages 3-4, the EchoRCNN including a regression network which predicts (generates) from input frames of a cardiac cycle video, offsets (difference) between A anchors from spatial positions of an object from a classification subnet (software-tracked contour), and a ground truth from a manually refined border (contour)), comprising:
training a first machine learning model with multiple first training data sets, each of the multiple first training data sets comprises a first training image set as the input for training, and a training difference parameter set as [the target for training] (Sirjani pages 3–4, fig. 1, training the network using LV and RV datasets, the datasets described on pages 2–3 as being 2D echocardiography frames from a video, and where one of the inputs to the segmentation subnet is the output of the regression subnet, described on page 4 as being a predicted offset (difference) between anchor points A, which are training difference parameters because they are offsets/differences determined during training), wherein the first training image set and the training difference parameter set are derived from a first training video (Sirjani page 2, right column, “Materials and methods” section, training of the EchoRNN using 750 selected echocardiography sequences as video with 45 frames on average, and where each video is processed one by one as shown in fig. 1, and disclosed on page 3, EchoRCNN architecture section) and generated by the steps of:
(a) obtaining the first training image set by selecting at least one image frame from the first training video (Sirjani page 2, collection of 2D echocardiography series were prepared from selecting 750 view series with 45 frames on average, which are selected videos (first training video) with proper LV (left ventricle) shapes, where fig. 1, and page 3, EchoRCNN architecture processing each frame of the video);
(b) generating, by an analysis software, a training software-tracked contour of the object from the first training video or the first training image set (Sirjani page 2, delineation (software-tracked contour) upon frames of each view series (first training video) was performed using the Auto 2D Quantification (a2DQ) tool in the Qlab Cardiac Analysis (analysis software));
(c) obtaining a training adjusted contour of the object (Sirjani page 2, users manually refine the points on the walls to correct the estimated region for LV (the object), the manual refined points used in the course of training the EchoRCNN and thus being a training adjusted contour); and
(d) obtaining the training difference parameter set based on the training software-tracked contour and the training adjusted contour (Sirjani page 4, the four outputs of the regression subnet is an offset (difference parameter) between A anchors from spatial positions of an object from a classification subnet (software-tracked contour), and a ground truth from a manually refined border (adjusted contour));
[wherein after training, the difference model uses the input image or the input video to generate the difference parameters without requiring the software-tracked contour or the adjusted contour].
While Sirjani teaches that offsets are predicted by a regression subnet and input into the segmentation subject, Sirjani does not explicitly teach that the offsets are “the target for training.”
Further, Sirjani does not explicitly teach “wherein after training, the difference model uses the input image or the input video to generate difference parameters without requiring the software-tracked contour or the adjusted contour.”
Isaksson teaches a difference parameter set as the target for training (Isaksson page 3, “Predicting contour quality” section, the target for training the contour quality model was two parameters that measure the difference between a machine determined contour and a ground truth manually generated contours, these two parameters being the mean absolute error and the Spearman rank correlation between a predicted Dice coefficient and the target Dice coefficient).
Isaksson further teaches wherein after training, the difference model uses the input image or the input video to generate the difference parameters without requiring the software-tracked contour or the adjusted contour (Isaksson page 9, Abstract, the deep learning model is trained so that in practice (after training – following the above cited “predicting contour quality” section which references fig. 1 as well), the model monitors the performance (by outputting a DICE score – generating difference parameters) using automated contour models, without the training data contours, thus without requiring the software-tracked contour or the adjusted contour, that is, as shown in fig. 1, outside of training, only an image is required to be provided for input).
Therefore, taking the teachings of Sirjani and Isaksson together as a whole, it would have been obvious to a person having ordinary skill in the art (herein “PHOSITA”) before the effective filing date of the claimed invention to have modified the regression subnet of Sirjani to include contour difference values as training targets and generating contour difference values in practice after training without needing the training data as disclosed in Isaksson, at least because doing so would allow for ensuring quality and monitoring the performance of deployed automated contouring models. See Isaksson Conclusions section on page 9.
Regarding claim 2, Sirjani teaches wherein each image in the first training image set is an echocardiographic image, and where the object is endocardium (Sirjani pages 2–4, training using LV and RV datasets from 2D echocardiography series, where LV (left ventricles) and RV (right ventricles) include respective endocardium).
Regarding claim 3, Sirjani teaches wherein each image of the first training image set is processed according to the software-tracked contour before used as the input for training (Sirjani pages 2–3, Fig. 1, delineation (software-tracked contour) was performed using the Auto 2D Quantification (a2DQ) tool as a first step before pre-processing and inputting into the main EchoRNN network).
Regarding claim 4, Sirjani teaches wherein the first machine learning model is a regression model based on convolutional neural network (Sirjani page 4, Figs. 1 and 2, regression subnet with convolutional layers like the classification subnet).
Regarding claim 5, Sirjani teaches wherein the first machine learning model is a residual neural network (ResNet) model (Sirjani Fig. 3, page 5, “box subnet” which is the regression subnet comprising a ResNet50 backbone).
Allowable Subject Matter
Claim 6, and claims 7–22 and 24–25 which depend therefrom are allowed. The following is a statement of reasons for the indication of allowable subject matter: the closest cited art of record includes Sirjani in combination with Isaksson, as applied above to claim 1. Further, Isaksson, while teaching “the software-generated analysis result is derived from at least one software-tracked contour of the object generated by the analysis software from the at least one input image or the input video; the adjusted analysis result is derived from at least one adjusted contour of the object in the at least one input image or the input video; and after training, the evaluation model uses at least one difference parameter set,” as given in pages 3–4 of Isaksson teaching evaluating the quality for automatically generated contours (software-tracked contour) against ground truth segmentations/contours determined by human experts (adjusted analysis result from at least one adjusted contour), and that after training the trained machine learning model in Isaksson (corresponding to the claimed evaluation model) uses a difference parameter set of a mean absolute error and Spearman rank of Dice coefficient values. Neither Isaksson, nor Sirjani, nor any of the other cited art of record, whether considered alone or in an obvious combination, teach or suggest to a person having ordinary skill in the art, the limitations recited in claim 6 of “after training, the evaluation model uses … and at least one geometric parameter set to generate the predicted evaluation errors, wherein each of the at least one geometric parameter set comprises one or more geometric parameters calculated based on one of the at least one software-tracked contour,” and all other supporting limitations thereof in claim 6.
Therefore, no combination of the cited art of record, whether considered alone, or in a combination obvious to a PHOSITA, teach or suggest the limitations of claim 6, and therefore claims 7–22, and 24–25 which presently recite a dependency from claim 6.
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 MICHELLE M KOETH whose telephone number is (571)272-5908. The examiner can normally be reached Monday-Thursday, 09:00-17:00, Friday 09:00-13:00, EDT/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, Vincent Rudolph can be reached at 571-272-8243. 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.
MICHELLE M. KOETH
Primary Examiner
Art Unit 2671
/MICHELLE M KOETH/Primary Examiner, Art Unit 2671
Prosecution Insights