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 filed February 20, 2026 have been fully considered and are in part persuasive and in part are not persuasive. Applicant argues that a newly-added limitation of claim 1 is not taught by Casale. In particular, Applicant argues that the following limitation is not taught or suggested by the Itu, Casale or Zaiss: “’modelling each of the plurality of sets of output embeddings output by one of the plurality of decoders in a probabilistic distribution using a respective Gaussian process’ as recited in amended claim 1.” In the previous Office Action, the examiner relied on Casale as disclosing “a Gaussian Process Prior Variational Encoder (GPPVAE) designed to model sets of output embeddings in probabilistic distributions, namely Gaussian distributions (pages 1-6, sections 1-3).”
Specifically, Applicant argues:
The cited portions of Casale disclose encoding sample images into a latent representation matrix Z in a low-dimensional space and decoding the latent representation matrix Z to generate images y in an original image space, where the latent representation matrix Z are modeled through a Gaussian process (GP) prior. However, modeling the latent representation matrix Z through a Gaussian process prior, as described in the cited portions of Casale, does not teach or suggest at least "modelling each of the plurality of sets of output embeddings output by one of the plurality of decoders in a probabilistic distribution using a respective Gaussian process" as recited in amended claim 1. Specifically, the latent representation matrix Z that is modeled through a Gaussian process prior in the cited portions of Casale is not "output embeddings output by one of the plurality of decoders" as recited in amended claim 1. Instead, the latent representation matrix Z is merely output by an encoder encoding the sample images. The cited portions of Casale do not teach or suggest at least "modelling each of the plurality of sets of output embeddings output by one of the plurality of decoders in a probabilistic distribution using a respective Gaussian process" as recited in amended claim 1.
The examiner agrees that Casale does not explicitly disclose modeling the output embeddings of the decoder in a probabilistic distribution because the probabilistic distribution modeling of Casale appears to occur in the encoder rather than following the decoder. Therefore, the rejection of independent claims 1, 10 and 15 is withdrawn. However, a new rejection of these claims is set forth below.
Claim Interpretation
The claims in this application are given their broadest reasonable interpretation (BRI) using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The BRI of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in this Office action. Claim limitations in this application that do not use the word “means” are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in this Office action.
Because claims 10 – 12 recite the word “means” or depend from a claim that recites the word “means”, they are presumed to invoke 35 U.S.C. 112(f). The presumption is overcome when the claim further includes structure necessary to perform the recited function. MPEP 2181(I) provides a 3-prong test for determining whether a “means” term in a claim invokes 35 U.S.C. 112(f):
(A) the claim limitation uses the term "means" or "step" or a term used as a substitute for "means" that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term "means" or "step" or the generic placeholder is modified by functional language, typically, but not always linked by the transition word "for" (e.g., "means for") or another linking word or phrase, such as "configured to" or "so that"; and
(C) the term "means" or "step" or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Regarding “means for receiving a temporal sequence of medical images of a vessel of a patient” in claim 10, this phrase clearly meets prongs A and B. Since the phrase does not recite any structure for performing the function of “receiving a temporal sequence of medical images of a vessel of a patient”, the phrase also meets prong C and thus invokes 35 U.S.C. 112(f). Once a determination is made that 35 U.S.C. 112(f) is invoked, the BRI for interpreting the claim is the structure, material or acts disclosed in the specification and equivalents thereof. Para. [0023] of the present disclosure describes the encoder 206 of the machine learning model as performing the function of receiving the temporal sequence. Therefore, the BRI for this claim element is the encoder 206 and equivalents thereof.
Regarding “means for generating a plurality of sets of output embeddings using a machine learning based model trained using multi-task learning” in claim 10, this phrase clearly meets prongs A and B. Since the phrase does not include the structure for performing the function, this phrase also meets prong C, and is therefore interpreted as invoking 35 U.S.C. 112(f). The BRI for this term is the structure disclosed in the present disclosure for performing this function, namely, the decoders 210, and equivalents thereof.
Regarding “means for modeling each of the plurality of sets…” in claim 10, this phrase clearly meets prongs A and B. Prong C is also met because the claim does not include the structure, material or acts for performing the function. Therefore, this phrase is interpreted as invoking 35 U.S.C. 112(f). Accordingly, the BRI for this term is the decoder structure 210 that models the sets of output embeddings in respective probabilistic distributions, and equivalents thereof.
Regarding “means for outputting the results of the plurality of vessel assessment tasks” in claim 10, this phrase clearly meets prongs A and B. Since the phrase does not recite any structure for performing the function of “outputting results”, the phrase also meets prong C and therefore invokes 35 U.S.C. 112(f). The BRI for this limitation is the structure, material or acts disclosed in the specification and equivalents thereof. Para. [0023] of the present disclosure refers to the decoders 210 shown in Fig. 2 as performing this function. Therefore, the BRI for this claim element is the decoders 210 and equivalents thereof.
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 1-3, 7-12, 15 and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Itu, U.S. Publ. Appl. No. 2021/0064936 A1 (hereinafter referred to as “Itu”) in view of Chinese Patent No. CN 114897748 A to Peng et al. (hereinafter referred to as “Peng”) and further in view of U.S. Publ. Appl. No. 2022/0179026 A1 of Zaiss, et al. (hereinafter referred to as “Zaiss”).
Regarding claim 1, Itu discloses a computer-implemented method (Fig. 6, computer 602, paras. [0071]-[0077]) comprising:
receiving a temporal sequence of medical images of a vessel of a patient (Fig. 1, step 102; Fig. 5, medical images 514-A – 514-C input to encoder portions 504/506; para. [0063] discusses leveraging temporal information from the sequence of input images; para. [0065] discusses the 2D CNN 504 and the LSTM RNN 506 forming the encoding portion of the machine learning model 502) that receives the input image sequence; para. [0036] discusses acquisition of “each temporal sequence” acquired by the image acquisition device);
generating a plurality of sets of output embeddings (Fig. 5, para. [0059], decoding portions formed by the 3D LTSM network 510 and decoder 512 generate output embeddings representing predictions of segmentation results) using a machine learning based model trained using multi-task learning (Fig. 5, the machine learning model 502 uses multi-task learning; para. [0049]), the machine learning based model comprising an encoder for encoding the temporal sequence of medical images into shared features and a plurality of decoders each for decoding the shared features into a respective one of the plurality of sets of output embeddings (para. [0061] discloses that the encoders 504-A – 504-C shown in Fig. 5 encode the temporal sequence of medical images into shared features 508-A – 508-C that constitute embeddings that are input to respective decoders 512-A – 512-C; para. [0063] discusses shared features such as varying sizes of the lumen at different times of the heart cycle extracted from the temporal sequence of images; para. [0022] discusses the temporal sequences of images that are acquired and input to the encoders of the model 502; as is known in the art, decoders either output vectors comprising embeddings/latent representations or include final layers that map the embeddings into another form, such as viewable images or text);
modeling each of a plurality of sets of output embeddings output by one of the plurality of decoders in a probabilistic distribution using a respective Gaussian process to generate, 1) results of a particular one of a plurality of vessel assessment tasks (para. [0025] of Itu discloses that the model 502 generates results 516-A -516-C of a particular one of a plurality of vessel assessment tasks, e.g., a prediction or quantification of coronary artery disease including fractional flow reserve (FFR), coronary flow reserve (CFR), instantaneous wave free ratio (iFR), basil stenosis resistance (BSR), etc., by modelling each of the plurality of sets of output embeddings in a machine learning model 502; para. [0059] discusses machine learning model 502 modelling the sets of output embeddings to perform the vessel assessment tasks; however, Itu does not disclose that the sets of output embeddings that are output from the decoders are modeled in respective Gaussian processes) and 2) a confidence measure associated with the results of the particular vessel assessment task, the confidence measure determined based on an entropy of the probabilistic distribution modelling the set of output embeddings (Itu does not explicitly disclose limitation 2)); and
outputting the results of the plurality of vessel assessment tasks and the confidence measure (Fig. 5, para. [0062], decoder 512 outputs the results of the plurality of the vessel assessment tasks).
As indicated above, Itu does not explicitly disclose that the machine learning model 502 models the sets of output embeddings that are output from each of the decoders 512-A – 512-C in a probabilistic distribution using a respective Gaussian process to generate the results.
Peng, in the same field of endeavor, discloses that the decoder performs upsampling on extracted deep features output from the encoder to obtain multi-level decoding feature embeddings that are output from the decoder. The multi-level decoding features are then input to a normalization layer that embeds a Gaussian distribution vector into the multi-level decoding feature to obtain a multi-level modal embedding feature. The multi-level modal embedding is then processed through additional layers along with a symmetric encoding feature to produce a missing modality image, i.e., to synthesize a medical imaging modality other than the one used to acquire the medical images. The modeling performed by the combination of the normalization layer and the layers that follow it constitute modelling the output embeddings of the decoder in a Gaussian process.
It would have been obvious to one of ordinary skill in the art to which the claimed invention pertains, before the effective filing date of the present disclosure, to use the Gaussian process modeling of Peng to model each of the sets of output embeddings 516-A – 516-C of Itu to generate the respective assessment task results. One of ordinary skill in the art would have been motivated to make the modification to allow the system and method of Itu to be used to obtain assessment task results for medical imaging modalities in addition to the modality that was used to acquire the images. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods to achieve predictable results (generating software that implements the normalization layer and the layers following the normalization layer such that the embeddings that are output from the decoders 512-A – 512-C are input to the normalization layer).
The combined teachings of Itu and Peng do not explicitly teach generating a confidence measure associated with the results of the particular vessel assessment task, the confidence measure determined based on an entropy of the probabilistic distribution modelling the respective set of output embeddings.
Zaiss, in the same field of endeavor, discloses determining a confidence measure based on an entropy of probabilistic distributions. The meaning of the claim language “based on an entropy of the probabilistic distributions” is not defined in the present disclosure. Therefore, the phrase is interpreted according to its plain meaning as understood by those of skill in the art. As indicated in the final rejection mailed June 4, 2025, the plain meaning of entropy of a probabilistic distribution is provided in “Lecture 6; Using Entropy for Evaluating and Comparing Probability Distributions Readings: Jurafsky and Martin, section 6.7”, University of Rochester, Department of Computer Science, posted on its website on January 26, 2021: “[t]he key concept used here is called entropy, which is a measure of the inherent randomness in a probability distribution (or set of observed data).”
Therefore, the BRI for the claim limitation “based on an entropy of the probabilistic distributions” is that it means “based on the inherent randomness of the probabilistic distributions”. In other words, entropy is an inherent property of a probabilistic distribution.
In Zaiss, the Gaussian distributions that are output from the machine learning model that processes the medical images have entropies that are inherent properties of the Gaussian distributions. The SD σ(x) parameters of the probabilistic distributions of Zaiss indicate the uncertainty and therefore constitute a confidence measure quantifying uncertainty for results based on an entropy of the probabilistic distributions (Zaiss, neural network 1 processes medical images and outputs Gaussian probability distributions having an SD σ(x) parameter that indicates the uncertainty of target parameters, para. [0089]).
Zaiss discloses that the uncertainty quantification measure can comprise an uncertainty map (paras. [0059]-[0061]: “at least one uncertainty quantification measure comprises an uncertainty map of an uncertainty quantity of the at least one image parameter”). Zaiss also discloses using probabilistic predictive models that generate both (1) a result of a vessel assessment task, referred to in Zaiss as an “image parameter of an image sample”, and (2) a measure of the uncertainty associated with the image parameter (para. [0051]: “[a]ccording to the invention, at least one image parameter of the sample and additionally at least one uncertainty quantification measure representing a prediction error of the at least one image parameter are provided by output elements of the neural network.”; para. [0060] discusses the image parameter being a “parameter map” of the image sample and the measure of uncertainty being an “uncertainty map”; paras. [0070]-[0071] discuss the different types of image parameters representing the vessel assessment task results; paras. [0065] and [0088] discusses the probablistic process configuration of the neural network providing the uncertainty measurements for the respective image parameters).
It would have been obvious to one of ordinary skill in the art to which the claimed invention pertains before the effective filing date of the present disclosure to use the probabilistic process of Zaiss, et al. in combination with the Gaussian process of Peng in the model 502 of Itu to generate a confidence measure such as an uncertainty map as taught by Zaiss and to output the uncertainty maps together with the respective results of the vessel assessment tasks as taught by Zaiss. One of ordinary skill in the art would have been motivated to make the modification to allow the user to observe the level of uncertainty associated with the vessel task assessment along with the vessel assessment task results. The modification could have been made by one of ordinary skill in the art before the effective filing date of the present disclosure with a reasonable expectation of success because making the modification merely involves combining prior art elements according to known methods (including software for implementing the model 502 that generates confidence measures in association with the assessment task results) to yield predictable results.
Regarding claim 2, Itu discloses that the temporal sequence of medical images is an angiography sequence (para. [0061], "[e]ach input medical image 514-A, 514-B, and 514-C comprises a sequence of images. For example, a respective input medical image 514 may comprise a plurality or all frames of coronary angiographies.").
Regarding claim 3, Itu discloses that the temporal sequence of medical images is acquired at a plurality of different acquisition angles (Fig. 5, 518; para. [0061], "view parameters 518 are features describing the input medical images 514, such as, e.g., C-arm angulation."; the C-arm angulation refers to the “angulation of the C-arm of the image acquisition device used to acquire each temporal sequence”, as described in para. [0036]).
Regarding claim 7, Itu discloses that the plurality of vessel assessment tasks comprises localization of a stenosis in the vessel (para. [0026] discusses that the one or more secondary tasks performed by the machine learning model include “predicting a location of stenosis markers in the input medical images”).
Regarding claim 8, Itu discloses that the plurality of vessel assessment tasks comprises image-based stenosis grading of a stenosis in the vessel (para. [0025] discusses the primary task being predicting a measure of the "significance of a stenosis in an artery", which means grading the stenosis; para. [0027] indicates that the stenosis grading is image-based because the predicted measures of interest for the primary and secondary task(s) may be “displayed along with the one or more medical images to facilitate user evaluation of the predicted measures of inter-est (e.g., for the one or more secondary tasks)”).
Regarding claim 9, Itu discloses that the plurality of vessel assessment tasks comprises labelling of a stenosis in the vessel and segmentation of the stenosis (para. [0026] discusses “predicting a segmentation of one, several, or all vessels visible in the input medical images"; Fig. 5 shows segmentation of vessels as outputs 516-A, 516-B and 516-C of the machine learning model 502, as described in para. [0059]). With respect to the limitation “labelling of a stenosis”, the BRI for this limitation is that it means classification of a stenosis because para. [0032] of the present disclosure uses the term “labelling” interchangeably with “classification”: “[i]n one embodiment, the plurality of vessel assessment tasks includes classification (i.e., labelling)….” Itu “classifies a stenosis” because it predicts a “location of stenosis markers” and quantifies the “significance of a stenosis”, as indicated in para. [0025], and displays the predictions “along with the one or more medical images to facilitate user evaluation of the predicted measures of inter-est (e.g., for the one or more secondary tasks)”, as indicated in para. [0027].
Regarding claims 10-12, the rejections of claims 1-3 apply mutatis mutandis to claims 10-12, respectively.
Regarding claims 15, to the extent that claim 15 recites the same limitations that are recited in claim 1, the rejection of claim 1 applies mutatis mutandis to claim 15. The only limitation that is recited in claim 15 that is not also recited in claim 1 is the non-transitory computer readable medium. Itu discloses that the machine learning model includes non-transitory computer readable medium (e.g., para. [0060] long short-term memory (LTSM)).
Regarding claims 18-20, the rejections of claims 7-9 apply mutatis mutandis to claims 18-20, respectively.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL J SANTOS whose telephone number is (571)272-2867. The examiner can normally be reached M-F 9-5.
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, Matt Bella, can be reached at (571)272-7778. 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.
/DANIEL J. SANTOS/ Examiner, Art Unit 2667
/MATTHEW C BELLA/ Supervisory Patent Examiner, Art Unit 2667
Prosecution Insights