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 .
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-4, and 18-21 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al (CN111134727B; hereinafter referred to as Srinivasan) in view of Qi et al (S. Qi, J. Li, and Z. Sun, “Adaptive confidence threshold algorithm for vehicle detection by employing temporal information,” 2018 10th International Conference on Intelligent Human-Machine Systems and Cybernetics (IHMSC), pp. 348–352, Aug. 2018; hereinafter referred to as Qi).
Regarding Claim 1, Li discloses an image processing apparatus (“The invention relates to the technical field of medical ultrasound, in particular to a vein and artery identification method and system based on a neural network.” [n0001]) comprising:
a processor configured to analyze each of a plurality of ultrasound images of a subject to detect a structure in each of the plurality of ultrasound images and calculate a certainty that the structure is a blood vessel for each of the plurality of ultrasound images (“ randomly selecting 3/5 of the images in all the collected ultrasound images as a training set; randomly selecting 1/5 of images as a verification set; the remaining 1/5 of the ultrasound images are used as a test set; the training set is used for training the neural network model, the verification set is used for verifying the recognition accuracy of the neural network and optimizing weight parameters of the neural network model, and the test set is used for finally evaluating the recognition accuracy of the neural network model; of course, the randomly selected proportion can be 3/5, 1/5, or other proportions;” [n0091], “inputting an ultrasonic image to be recognized into a neural network model obtained through training for processing; acquiring position information of veins and arteries from an ultrasonic image to be identified through the neural network model; the marked veins and arteries are distinguished according to the acquired position information, and an ultrasound image containing the vein marks and the artery marks is generated.” [n0085],
the certainty being an index representing a likelihood that the structure is the blood vessel, the certainty being represented by a probability that the structure is the blood vessel (“ the information of the bounding box comprises probability information that the image in the bounding box is an artery or a vein, and position information and size information of the bounding box.” [n0051] “information of each boundary frame needs to be represented by 2 + 4 + 1 = 7 number, wherein 2 numbers respectively represent the image in the boundary frame is artery, probability information of vein, two probability information are respectively recorded as c1, c2; 4 numbers represent the coordinate information of the central position of the boundary frame, namely abscissa, ordinate and length, width information; 1 number records the size of the possibility of the artery or vein contained in the boundary frame; at last setting softmax classification layer of the neural network model, limiting 2 probability information to 0 to 1, and when the boundary frame contains artery or vein, 2 probability information c1, c2 sum is 1;” [n0110]),
detect, as the blood vessel, the structure having the certainty higher than a certainty threshold value (“Further, screening boundary frame refers to selecting the prediction probability is greater than the set threshold of the boundary frame as the prediction result; in the boundary frame with prediction probability greater than the set threshold, using maximum suppression method for further screening, the specific method is calculating the overlapping degree between the boundary frame, the overlapping degree index is greater than the boundary frame of the set threshold value, selecting the boundary frame with the highest prediction probability as the identification result” [n0141]),
Li does not specifically disclose changing the certainty threshold value based on a plurality of the calculated certainties that the structure in the plurality of images is the desired structure, wherein the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated.
However, in a similar field of endeavor, Qi teaches an adaptive confidence algorithm for vehicle detection using convolutional neural networks is proposed [Abstract].
Qi also teaches changing the certainty threshold value based on a plurality of the calculated certainties that the structure in the plurality of images is the desired structure, wherein the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated (“There exists two problems in vehicle detection– false alarms and false negatives. When a false alarm appears in the current frame, it must be that its confidence is larger than fixed confidence thresh we set in advance. In this situation, we should increase confidence thresh to ensure it can not be selected as a vehicle. When a vehicle does not appear in the current frame, it must be that its confidence is smaller than fixed confidence thresh. Now, we should decrease confidence thresh to ensure it can be selected as a vehicle. To solve this problem, we introduce parameter K to learn an optimal confidence. This is achieved with the following equation. N ST =FT− i=1 (Ci −FT)e−Ki (1) where ST ∈ (0,1) is the dynamic confidence that we deter mine, i is the order of the pictures before current frame, N is the number of vehicles that are regarded as relevant to current vehicle, FT is fixed confidence thresh in the MS CNN algorithm, and Ci is the confidence of the front frame, if it does not exist corresponding ROI, Ci is zero.” [Pg. 349 Col. 2 Lines 3-20]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li as outlined above with changing the certainty threshold value based on a plurality of the calculated certainties that the structure in the plurality of images is the desired structure, wherein the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated as taught by Qi, because It can effectively reduce the false alarm rate and increase the true positive rate [Pg. 348 Col. 2 Lines 23-25].
Regarding Claim 2, Li discloses that the apparatus further comprising: an ultrasound probe; and wherein the processor is further configured to generate the plurality of ultrasound images based on transmission and reception of an ultrasound beam using the ultrasound probe (“Fig. 1-a shows a neural network based real-time ultrasonic guidance system for jugular vein puncture, which comprises: the transducer is used for transmitting and receiving ultrasonic signals; ” [n0074]).
Regarding Claim 3, Li discloses that further comprising: a certainty memory, wherein the processor is further configured to store the certainty in the certainty memory (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [n0070]).
Regarding Claim 4, Li discloses that the processor is further configured to control storage of the certainty in the certainty memory (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [n0070]).
Regarding Claim 18, Srinivasan discloses that further comprising: a certainty threshold memory; the processor is further configured to store the changed certainty threshold value in the certainty threshold memory for each subject (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [Li n0070]).
Regarding Claim 19, Srinivasan discloses that the apparatus further comprises: an input device configured to accept an input operation of a user wherein the processor is further configured to change the certainty threshold value based on the input operation via the input device (“inputting an ultrasonic image to be identified into a trained neural network model, and acquiring all bounding boxes output by the neural network model; and screening the output bounding box according to a set probability threshold value so as to obtain the position information of the vein and the artery. Further, the bounding box that is output by screening according to the set probability threshold specifically includes: selecting a bounding box with the prediction probability larger than a set probability threshold value as a prediction result; and in the boundary frames with the prediction probability larger than the set probability threshold, screening the boundary frame with the highest prediction probability by adopting a maximum suppression method as a screening result, and further acquiring the position information of the veins and the arteries.” [0034-0037]).
Regarding Claim 20, Li discloses a control method of an image processing apparatus (“The invention relates to the technical field of medical ultrasound, in particular to a vein and artery identification method and system based on a neural network.” [n0001]) comprising:
analyzing each of a plurality of ultrasound images of a subject to detect a structure in each of the plurality of ultrasound images and calculate a certainty that the structure is a blood vessel for each of the plurality of ultrasound images (“ randomly selecting 3/5 of the images in all the collected ultrasound images as a training set; randomly selecting 1/5 of images as a verification set; the remaining 1/5 of the ultrasound images are used as a test set; the training set is used for training the neural network model, the verification set is used for verifying the recognition accuracy of the neural network and optimizing weight parameters of the neural network model, and the test set is used for finally evaluating the recognition accuracy of the neural network model; of course, the randomly selected proportion can be 3/5, 1/5, or other proportions;” [n0091], “inputting an ultrasonic image to be recognized into a neural network model obtained through training for processing; acquiring position information of veins and arteries from an ultrasonic image to be identified through the neural network model; the marked veins and arteries are distinguished according to the acquired position information, and an ultrasound image containing the vein marks and the artery marks is generated.” [n0085],
the certainty being an index representing a likelihood that the structure is the blood vessel, the certainty being represented by a probability that the structure is the blood vessel (“ the information of the bounding box comprises probability information that the image in the bounding box is an artery or a vein, and position information and size information of the bounding box.” [n0051] “information of each boundary frame needs to be represented by 2 + 4 + 1 = 7 number, wherein 2 numbers respectively represent the image in the boundary frame is artery, probability information of vein, two probability information are respectively recorded as c1, c2; 4 numbers represent the coordinate information of the central position of the boundary frame, namely abscissa, ordinate and length, width information; 1 number records the size of the possibility of the artery or vein contained in the boundary frame; at last setting softmax classification layer of the neural network model, limiting 2 probability information to 0 to 1, and when the boundary frame contains artery or vein, 2 probability information c1, c2 sum is 1;” [n0110]),
detecting, as the blood vessel, the structure having the certainty higher than a certainty threshold value (“Further, screening boundary frame refers to selecting the prediction probability is greater than the set threshold of the boundary frame as the prediction result; in the boundary frame with prediction probability greater than the set threshold, using maximum suppression method for further screening, the specific method is calculating the overlapping degree between the boundary frame, the overlapping degree index is greater than the boundary frame of the set threshold value, selecting the boundary frame with the highest prediction probability as the identification result” [n0141]),
Li does not specifically disclose changing the certainty threshold value based on a plurality of the calculated certainties that the structure in the plurality of images is the desired structure, wherein the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated.
However, in a similar field of endeavor, Qi teaches an adaptive confidence algorithm for vehicle detection using convolutional neural networks is proposed [Abstract].
Qi also teaches changing the certainty threshold value based on a plurality of the calculated certainties that the structure in the plurality of images is the desired structure, wherein the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated (“There exists two problems in vehicle detection– false alarms and false negatives. When a false alarm appears in the current frame, it must be that its confidence is larger than fixed confidence thresh we set in advance. In this situation, we should increase confidence thresh to ensure it can not be selected as a vehicle. When a vehicle does not appear in the current frame, it must be that its confidence is smaller than fixed confidence thresh. Now, we should decrease confidence thresh to ensure it can be selected as a vehicle. To solve this problem, we introduce parameter K to learn an optimal confidence. This is achieved with the following equation. N ST =FT− i=1 (Ci −FT)e−Ki (1) where ST ∈ (0,1) is the dynamic confidence that we deter mine, i is the order of the pictures before current frame, N is the number of vehicles that are regarded as relevant to current vehicle, FT is fixed confidence thresh in the MS CNN algorithm, and Ci is the confidence of the front frame, if it does not exist corresponding ROI, Ci is zero.” [Pg. 349 Col. 2 Lines 3-20]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li as outlined above with changing the certainty threshold value based on a plurality of the calculated certainties that the structure in the plurality of images is the desired structure, wherein the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated as taught by Qi, because It can effectively reduce the false alarm rate and increase the true positive rate [Pg. 348 Col. 2 Lines 23-25].
Regarding Claim 21, Li discloses that the certainty is calculated using machine learning (“inputting an ultrasonic image to be identified into a trained neural network model, and acquiring all bounding boxes output by the neural network model; and screening the output bounding box according to a set probability threshold value so as to obtain the position information of the vein and the artery. Further, the bounding box that is output by screening according to the set probability threshold specifically includes: selecting a bounding box with the prediction probability larger than a set probability threshold value as a prediction result; and in the boundary frames with the prediction probability larger than the set probability threshold, screening the boundary frame with the highest prediction probability by adopting a maximum suppression method as a screening result, and further acquiring the position information of the veins and the arteries.” [0034-0037]).
Claims 5-6 is rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Qi as applied to Claim 4 above, and further in view of Radulescu et al (US20200043129A1; hereinafter referred to as Radulescu).
Regarding Claim 5, Li in view of Qi discloses that the processor is further configured to store the certainty in the certainty memory (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [Li n0070]).
Li in view of Qi does not specifically disclose that the processor is further configured determine that the ultrasound probe is stationary and store information in a case in which it is determined that the ultrasound probe is stationary for a determined time or longer.
However, in a similar field of endeavor, Radulescu teaches an apparatus includes an imaging probe and is configured for dynamically arranging presentation of visual feedback for guiding manual adjustment [Abstract].
Radulescu also teaches that the processor is further configured determine that the ultrasound probe is stationary and store information in a case in which it is determined that the ultrasound probe is stationary for a determined time or longer (“issuing a user alert for halting the adjustment is responsive to content of imaging dynamically acquired via the probe.” [0026], “the apparatus is configured for detecting the halting” [0027], “In a yet, further sub-aspect, the apparatus is further configured for, responsive to detecting that the halting has occurred, performing the segmenting.” [0028], “Once the movement halts (step S260), query is made, as in step S249, as to whether the current view is sufficiently on target for commencing quantification and optionally live imaging acquisition for storage (step S264)… the current view is sufficiently on target for commencing quantification and optionally live imaging acquisition for storage (step S264)” [0069]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor is further configured determine that the ultrasound probe is stationary and store information in a case in which it is determined that the ultrasound probe is stationary for a determined time or longer as taught by Radulescu, because would improve patient treatment [0007].
Regarding Claim 6, Li in view of Qi discloses that the processor is further configured to store the certainty in the certainty memory (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [Li n0070]).
Li in view of Qi does not specifically disclose that the processor is further configured to determine that the ultrasound probe is in contact with a body surface of the subject.
However, in a similar field of endeavor, Radulescu teaches that the processor is further configured to determine that the ultrasound probe is in contact with a body surface of the subject (“issuing a user alert for halting the adjustment is responsive to content of imaging dynamically acquired via the probe.” [0026], “the apparatus is configured for detecting the halting” [0027], “In a yet, further sub-aspect, the apparatus is further configured for, responsive to detecting that the halting has occurred, performing the segmenting.” [0028], “Once the movement halts (step S260), query is made, as in step S249, as to whether the current view is sufficiently on target for commencing quantification and optionally live imaging acquisition for storage (step S264)… the current view is sufficiently on target for commencing quantification and optionally live imaging acquisition for storage (step S264)” [0069]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor is further configured to determine that the ultrasound probe is in contact with a body surface of the subject as taught by Radulescu, because would improve patient treatment [0007].
Claims 7-8, & 17 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Qi as applied to Claims 1 and 4 above, and further in view of Matsumoto (US20190200963A1).
Regarding Claim 7, Li in view of Qi discloses that the processor is further configured to store the certainty in the certainty memory (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [Li n0070]).
Li in view of Qi does not specifically disclose that it is determined that a moving speed of the ultrasound probe is lower than a determined moving speed.
However, in a similar field of endeavor, Matsumoto teaches an ultrasound diagnostic apparatus and a control method of the ultrasound diagnostic apparatus [0002].
Matsumoto also teaches that it is determined that a moving speed of the ultrasound probe is lower than a determined moving speed (“In step S10, the image analyzing unit 13 performs image analysis on the ultrasound image further generated by the B-mode processing unit 20 of the image generating unit 6 in order to identify the part of the subject included in the ultrasound image. For example, in a case where a high frame rate at the time of the ultrasound diagnosis is set as the second imaging condition, the image analyzing unit 13 can perform optical flow, which is time-series image analysis, as the image analysis. Although not illustrated, the optical flow is a technique for mapping a direction and distance of movement of each pattern by using a vector or the like for a plurality of characteristic patterns in the same part included in ultrasound images of a plurality of frames in common by using the ultrasound images of a plurality of frames acquired by the image acquiring unit 3 in a time-series manner. By using such a time-series analysis method, for example, it can become easier to identify a part with much movement such as the heart and a part with a little movement such as the abdomen.” [0064]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with it being determined that a moving speed of the ultrasound probe is lower than a determined moving speed as taught by Matsumoto, because it can become easier to identify a part [0064].
Regarding Claim 8, Li in view of Qi discloses that the processor is further configured to store the certainty in the certainty memory (“The module can be configured to reside in addressable storage media and to execute on one or more processors. Modules can include components (such as software components, object-oriented software components, class components, and task components), processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables” [Li n0070]).
Li in view of Qi does not specifically disclose that the certainty being calculated based on the ultrasound image having a frame selected by a frame interval according to a moving speed of the ultrasound probe in the ultrasound image having a plurality of frames.
However, in a similar field of endeavor, Matsumoto teaches that the certainty being calculated based on the ultrasound image having a frame selected by a frame interval according to a moving speed of the ultrasound probe in the ultrasound image having a plurality of frames (“a part probability calculating unit that calculates, for the ultrasound image acquired in the image acquiring unit in accordance with a first imaging condition, a probability that a part of the subject included in the ultrasound image is a specific part from at least one of an orientation angle of the ultrasound probe or an analysis result of the ultrasound image;” [0008], “In step S10, the image analyzing unit 13 performs image analysis on the ultrasound image further generated by the B-mode processing unit 20 of the image generating unit 6 in order to identify the part of the subject included in the ultrasound image. For example, in a case where a high frame rate at the time of the ultrasound diagnosis is set as the second imaging condition, the image analyzing unit 13 can perform optical flow, which is time-series image analysis, as the image analysis. Although not illustrated, the optical flow is a technique for mapping a direction and distance of movement of each pattern by using a vector or the like for a plurality of characteristic patterns in the same part included in ultrasound images of a plurality of frames in common by using the ultrasound images of a plurality of frames acquired by the image acquiring unit 3 in a time-series manner. By using such a time-series analysis method, for example, it can become easier to identify a part with much movement such as the heart and a part with a little movement such as the abdomen.” [0064]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the certainty being calculated based on the ultrasound image having a frame selected by a frame interval according to a moving speed of the ultrasound probe in the ultrasound image having a plurality of frames as taught by Matsumoto, because it can become easier to identify a part [0064].
Regarding Claim 17, Li in view of Qi discloses all limitation noted above except that the processor is further configured to notify a user of a change of the certainty threshold value.
However, in a similar field of endeavor, Matsumoto teaches that the processor is further configured to notify a user of a change of the certainty threshold value (“Subsequently, when the number of times of determination that the plurality of part probabilities are all less than the threshold value becomes N, the process proceeds to st+ep S17. A message indicating that an error has occurred is displayed in the display unit 8, and then the ultrasound diagnostic apparatus 1 terminates the operation.” [0060]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor being further configured to notify a user of a change of the certainty threshold value as taught by Matsumoto, because it can indicate that an error has occurred [0060].
Claims 9-10 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Qi as applied to Claims 1, and further in view of Zou et al (Y. Zou, Z. Yu, B. V. Vijaya Kumar, and J. Wang, “Unsupervised domain adaptation for semantic segmentation via class-balanced self-training,” Lecture Notes in Computer Science, pp. 297–313, 2018; hereinafter referred to as Zou)
Regarding Claim 9, Li in view of Qi discloses all limitation noted above except that the processor is further configured to sort the plurality of the calculated certainties in ascending order or descending order, select, as the change value, one of the plurality of the calculated certainties located at a predetermined rank, and change the certainty threshold value to the change value.
However, in a similar field of endeavor, Zou teaches a novel UDA framework based on an iterative self-training procedure, where the problem is formulated as latent variable loss minimization, and can be solved by alternatively generating pseudo labels on target data and re-training the model with these labels [Abstract].
Zou also teaches that the processor is further configured to sort the plurality of the calculated certainties in ascending order or descending order, select, as the change value, one of the plurality of the calculated certainties located at a predetermined rank, and change the certainty threshold value to the change value (“Determination of k in ST From the previous section, we know that k essentially plays a key role in pseudo-label selection, by filtering out those with confidence lower than exp(−k). We can design the following policy on k such that we gradually increase the proportion of selected pseudo-labels in each round: We take the confidence of all pixels from the whole target set, and sort their confidence in a descending order. We then set the value of k such that exp(−k) equals to the probability ranked at round(p∗T∗N), where p indicates the pseudo label proportion and is between [0,1]. In this case, pseudo-label optimization produces p×100% most confident pseudo-labels for network training. The above pseudo-label selection policy is summarized in Algorithm 1... The policy of kc in CBST can be defined similar to ST. Although CBST seemingly introduce much more parameters than ST, we propose a strategy to determine kc for every class with a single param eter p, while effectively encoding the class-wise confidence levels. The proposed strategy is described in Algorithm 2. Note that Algorithm 2 determines kc by ranking the class c probabilities on all pixels predicted as class c, and setting kc such that exp(−kc) equals to the probability ranked at round(p ∗ Nc), where Nc indicates the number of pixels predicted as class c. Such a strategy basically takes the probability ranked at p × 100% separately from each class as a reference for both thresholding and confidence normalization. The pseudo-label proportion p and its increasing policy is defined exactly the same to ST” [Pg. 9]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor is further configured to sort the plurality of the calculated certainties in ascending order or descending order, select, as the change value, one of the plurality of the calculated certainties located at a predetermined rank, and change the certainty threshold value to the change value as taught by Zou, because it can refine the generated labels [Abstract].
Regarding Claim 10, Li in view of Qi discloses all limitation noted above except that the processor is further configured to sort the plurality of the calculated certainties in ascending order or descending order, select, as the change value, one of the plurality of the calculated certainties located at a rank corresponding to a predetermined ratio relative to a total number of the plurality of the calculated certainties, and change the certainty threshold value to the change value.
However, in a similar field of endeavor, Zou teaches that the processor is further configured to sort the plurality of the calculated certainties in ascending order or descending order, select, as the change value, one of the plurality of the calculated certainties located at a predetermined rank, and change the certainty threshold value to the change value (“Determination of k in ST From the previous section, we know that k essentially plays a key role in pseudo-label selection, by filtering out those with confidence lower than exp(−k). We can design the following policy on k such that we gradually increase the proportion of selected pseudo-labels in each round: We take the confidence of all pixels from the whole target set, and sort their confidence in a descending order. We then set the value of k such that exp(−k) equals to the probability ranked at round(p∗T∗N), where p indicates the pseudo label proportion and is between [0,1]. In this case, pseudo-label optimization produces p×100% most confident pseudo-labels for network training. The above pseudo-label selection policy is summarized in Algorithm 1... The policy of kc in CBST can be defined similar to ST. Although CBST seemingly introduce much more parameters than ST, we propose a strategy to determine kc for every class with a single param eter p, while effectively encoding the class-wise confidence levels. The proposed strategy is described in Algorithm 2. Note that Algorithm 2 determines kc by ranking the class c probabilities on all pixels predicted as class c, and setting kc such that exp(−kc) equals to the probability ranked at round(p ∗ Nc), where Nc indicates the number of pixels predicted as class c. Such a strategy basically takes the probability ranked at p × 100% separately from each class as a reference for both thresholding and confidence normalization. The pseudo-label proportion p and its increasing policy is defined exactly the same to ST” [Pg. 9]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor is further configured to sort the plurality of the calculated certainties in ascending order or descending order, select, as the change value, one of the plurality of the calculated certainties located at a predetermined rank, and change the certainty threshold value to the change value as taught by Zou, because it can refine the generated labels [Abstract].
Claims 11 & 16 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Qi as applied to Claims 1, and further in view of Yoon et al (Y. Yoon, A. Boragule, Y. Song, K. Yoon, and M. Jeon, “Online multi-object tracking with historical appearance matching and scene adaptive detection filtering,” 2018 15th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), pp. 1–6, Nov. 2018; hereinafter referred to as Yoon)
Regarding Claim 11, Li in view of Qi discloses all limitation noted above except that the processor is further configured to calculate, as the change value, a sum of an average value of the plurality of the calculated certainties and a constant multiple of a standard deviation of the plurality of the calculated certainties, and change the certainty threshold value to the change value.
However, in a similar field of endeavor, Yoon teaches methods to handle temporal errors during multi-object tracking. [Abstract].
Yoon also teaches that the processor is further configured to calculate, as the change value, a sum of an average value of the plurality of the calculated certainties and a constant multiple of a standard deviation of the plurality of the calculated certainties, and change the certainty threshold value to the change value (“Two cumulative distribution functions(P(D ≤ τ)) of the gaussian variable(D) are combined through β. The gaus sian variable(D = N(µ,σ)) is derived from the average(µ) and standard deviation(σ) of detection confidences. D10 t is calculated by detection confidences of recent 10 frames. Because of the reason that D10 t is usually calculated by the small number of samples, Dall t , calculated by all detection confidences collected until the current frame, is needed for smoothness.” [Pg. 4]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor is further configured to calculate, as the change value, a sum of an average value of the plurality of the calculated certainties and a constant multiple of a standard deviation of the plurality of the calculated certainties, and change the certainty threshold value to the change value as taught by Yoon, because tracking performance, especially identity consistency, is highly improved [Abstract].
Regarding Claim 16, Li discloses all limitations noted above except that the processor is further configured to change the certainty threshold value to the change value in a case in which the change value is lower than the certainty threshold value.
However, in a similar field of endeavor, Qi teaches the processor is further configured to change the certainty threshold value to the change value in a case in which the change value is lower than the certainty threshold value (“There exists two problems in vehicle detection– false alarms and false negatives. When a false alarm appears in the current frame, it must be that its confidence is larger than fixed confidence thresh we set in advance. In this situation, we should increase confidence thresh to ensure it can not be selected as a vehicle. When a vehicle does not appear in the current frame, it must be that its confidence is smaller than fixed confidence thresh. Now, we should decrease confidence thresh to ensure it can be selected as a vehicle. To solve this problem, we introduce parameter K to learn an optimal confidence. This is achieved with the following equation. N ST =FT− i=1 (Ci −FT)e−Ki (1) where ST ∈ (0,1) is the dynamic confidence that we deter mine, i is the order of the pictures before current frame, N is the number of vehicles that are regarded as relevant to current vehicle, FT is fixed confidence thresh in the MS CNN algorithm, and Ci is the confidence of the front frame, if it does not exist corresponding ROI, Ci is zero.” [Pg. 349 Col. 2 Lines 3-20]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li as outlined above with the processor is further configured to change the certainty threshold value to the change value in a case in which the change value is lower than the certainty threshold value as taught by Qi, because It can effectively reduce the false alarm rate and increase the true positive rate [Pg. 348 Col. 2 Lines 23-25].
Claims 12 & 15 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Qi as applied to Claims 1, and further in view of Zhou et al (B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba, “Learning deep features for discriminative localization,” 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2921–2929, Jun. 2016; hereinafter referred to as Zhou)
Regarding Claim 12, Li in view of Qi discloses all limitation noted above except that the processor is further configured to calculate the change value by multiplying a maximum value of the plurality of the calculated certainties by a determined ratio smaller than 1.0, and change the certainty threshold value to the change value.
However, in a similar field of endeavor, Zhou teaches procedure for generating class activation maps (CAM) using global average pooling (GAP) in convolutional neural networks (CNNs). [Pg. 2].
Zhou also teaches that the processor is further configured to calculate the change value by multiplying a maximum value of the plurality of the calculated certainties by a determined ratio smaller than 1.0, and change the certainty threshold value to the change value (“In order to perform localization, we need to generate a bounding box and its associated object cate gory. To generate a bounding box from the CAMs, we use a simple thresholding technique to segment the heatmap. We first segment the regions of which the value is above 20% of the max value of the CAM. Then we take the bounding box that covers the largest connected component in the seg mentation map. We do this for each of the top-5 predicted classes for the top-5 localization evaluation metric. Fig. 6(a) shows some example bounding boxes generated using this technique. The localization performance on the ILSVRC validation set is shown in Tbl. 2, and example outputs in Fig. 5” [Pg. 4-5]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li in view of Qi as outlined above with the processor is further configured to calculate the change value by multiplying a maximum value of the plurality of the calculated certainties by a determined ratio smaller than 1.0, and change the certainty threshold value to the change value as taught by Zhou, because it achieves the lowest localization error [Pg. 5].
Regarding Claim 15, Li discloses all limitations noted above except that the processor is further configured to change the certainty threshold value to the change value in a case in which the change value is lower than the certainty threshold value.
However, in a similar field of endeavor, Qi teaches the processor is further configured to change the certainty threshold value to the change value in a case in which the change value is lower than the certainty threshold value (“There exists two problems in vehicle detection– false alarms and false negatives. When a false alarm appears in the current frame, it must be that its confidence is larger than fixed confidence thresh we set in advance. In this situation, we should increase confidence thresh to ensure it can not be selected as a vehicle. When a vehicle does not appear in the current frame, it must be that its confidence is smaller than fixed confidence thresh. Now, we should decrease confidence thresh to ensure it can be selected as a vehicle. To solve this problem, we introduce parameter K to learn an optimal confidence. This is achieved with the following equation. N ST =FT− i=1 (Ci −FT)e−Ki (1) where ST ∈ (0,1) is the dynamic confidence that we deter mine, i is the order of the pictures before current frame, N is the number of vehicles that are regarded as relevant to current vehicle, FT is fixed confidence thresh in the MS CNN algorithm, and Ci is the confidence of the front frame, if it does not exist corresponding ROI, Ci is zero.” [Pg. 349 Col. 2 Lines 3-20]).
It would have been obvious to an ordinary skilled person in the art before the effective filing
date of the claimed invention to modify the system of Li as outlined above with the processor is further configured to change the certainty threshold value to the change value in a case in which the change value is lower than the certainty threshold value as taught by Qi, because It can effectively reduce the false alarm rate and increase the true positive rate [Pg. 348 Col. 2 Lines 23-25].
Response to Arguments
Applicant's arguments filed 08/10/2026 have been fully considered but they are not persuasive.
Regarding the U.S.C. 103 rejection of claims 1 & 20, the applicant argues the following:
Applicant contends that Li does not disclose or suggest changing the probability/certainty
threshold by calculating a change value through statistical analysis of a distribution of values of
calculated blood-vessel certainties from a plurality of ultrasound images.
Applicant contends that Qi does not cure this deficiency. Qi calculates a dynamic
confidence threshold ST from a fixed threshold FT, previous-frame confidence values Ci,
exponential weighting terms, and a parameter K. Qi is directed to vehicle-detection false alarms
and lost vehicles based on temporal correspondence of vehicle bounding boxes.
The amended claims do not merely recite adaptive threshold changing in the abstract. They
recite calculating a change value by statistically analyzing a distribution of values of calculated
certainties that a structure in ultrasound images is a blood vessel.
Even if Li and Qi were combined, Applicant contends that the combination would at most
suggest applying Qi's dynamic threshold formula to Li's vessel-recognition system. Applicant
contends that such a combination would not disclose or suggest the amended distribution-based
calculation of a change value from blood-vessel certainties calculated for ultrasound images.
In response to applicant's arguments against the references individually, one cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986).
However, it is noted that under broadest reasonable interpretation of claim 1, the newly amended claim only requires the threshold being changed by analyzing a distribution of values. A statistical analysis is recited highly broadly in the independent claim.
In the instant examination, Li provides a plurality of ultrasound images with a per image certainty of a blood vessel being detected as a probability as well as threshold comparison to ensure all blood vessels in the image are detected (“ randomly selecting 3/5 of the images in all the collected ultrasound images as a training set; randomly selecting 1/5 of images as a verification set; the remaining 1/5 of the ultrasound images are used as a test set; the training set is used for training the neural network model, the verification set is used for verifying the recognition accuracy of the neural network and optimizing weight parameters of the neural network model, and the test set is used for finally evaluating the recognition accuracy of the neural network model; of course, the randomly selected proportion can be 3/5, 1/5, or other proportions;” [n0091], “inputting an ultrasonic image to be recognized into a neural network model obtained through training for processing; acquiring position information of veins and arteries from an ultrasonic image to be identified through the neural network model; the marked veins and arteries are distinguished according to the acquired position information, and an ultrasound image containing the vein marks and the artery marks is generated.” [n0085]).
Qi provides an adaptive thresholding algorithm which uses a weighted sum (statistical analysis) of the current certainty value in comparison with previously calculated certainties (distribution of values) to determine a dynamic threshold (“This is achieved with the following equation, where ST ∈ (0,1) is the dynamic confidence that we deter mine, i is the order of the pictures before current frame, N is the number of vehicles that are regarded as relevant to current vehicle, FT is fixed confidence thresh in the MS CNN algorithm, and Ci is the confidence of the front frame, if it does not exist corresponding ROI, Ci is zero. When N is larger than 3, the increment of ST will not exceed e−4K =0.01831K. So N is set to 3 to ensure the increment of ST larger as soon as possible in our experiment.” [Pg. 349 Col. 2 Lines 3-20], Eqn 1 is provided
PNG
media_image1.png
62
266
media_image1.png
Greyscale
, it is also noted that Eqn 2 highlight how the K term is chosen to).
In combination both references broadly teach that the certainty threshold value is changed to a change value that is calculated by statistically analyzing a distribution of values of the plurality of the calculated
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 STEVEN MALDONADO whose telephone number is 703-756-1421. The examiner can normally be reached 8:00 am-4:00 pm PST M-Th 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, Christopher Koharski can be reached on (571) 272-7230. 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.
/Steven Maldonado/
Patent Examiner, Art Unit 3797
/CHRISTOPHER KOHARSKI/Supervisory Patent Examiner, Art Unit 3797
