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 Objections
Claims 1, 13, and 16 are objected to because of the following informalities: claims 1, 13, and 16 recite the limitation “dynamic positron emission tomography (dPET) data of the selected anatomy for the subject, in the time domain, with a tracer applied to the selected anatomy of the subject” should read “dynamic positron emission tomography (dPET) data of the selected anatomy for the subject, in a time domain, with a tracer applied to the selected anatomy of the subject”. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-16 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites the limitation "the co-registered MRI data frames" in line 11. There is insufficient antecedent basis for this limitation in the claim.
Claim 13 recites the limitation "the co-registered MRI data frames" in line 11. There is insufficient antecedent basis for this limitation in the claim.
Claim 16 recites the limitation "the co-registered MRI data frames" in line 10. There is insufficient antecedent basis for this limitation in the claim.
The rest of the claims are rejected based on their dependency.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-6 are rejected under 35 U.S.C. 103 as being unpatentable over Quigg et al. (WO 2022/132772, however the US 2024/0057950 version is used for examination purposes) in the view of Feng et al. (CN 117156078).
Regarding claim 1, Quigg teaches a computer implemented method of using digital images for mapping metabolic activity within a selected anatomy of a subject, the method comprising (paras. 0029 and 0034; dynamic PET processing can be conducted using one or more host computing devices. For example, FIG. 2 is a block diagram of an example computer platform system 200 for performing dynamic PET and associated processes. the disclosed subject matter relates to an alternative process of metabolic mapping, dynamic PET. This novel technology quantifies the dynamics of radiotracer uptake and decay in order to provide additional specificity or sensitivity in mapping of the desired metabolic activity in the target site (e.g., target organ).):
using a computer having a processor connected to computer memory storing software that, when executed, performs computer instruction steps of a machine learning architecture comprising (para. 0035; system 200 may include one or more computing platform(s) 202 (e.g., a dynamic PET platform) having one or more processor(s) 204, such as a central processing unit (e.g., a single core or multiple processing cores), a microprocessor, a microcontroller, a network processor, an application-specific integrated circuit (ASIC), or the like. Computing platform 202 may also include memory 206. Memory 206 may comprise random access memory (RAM), flash memory, a magnetic disk storage drive, and the like. In some embodiments, memory 206 may be configured to store a dynamic PET engine 208 and a trained artificial neural network (ANN) model 210 (e.g., a dynamic PET ANN).):
collecting magnetic resonance image (MRI) data of the selected anatomy of the subject including a preferred anatomical portion of the selected anatomy (para. 0036; dynamic PET engine 208 may be configured to receive image data from each of a PET scanner device 220 and/or a MRI scanner device 222.);
collecting dynamic positron emission tomography (dPET) data of the selected anatomy for the subject, in the time domain, with a tracer applied to the selected anatomy of the subject (para. 0036; dynamic PET engine 208 may be configured to receive image data from each of a PET scanner device 220 and/or a MRI scanner device 222.);
co-registering MRI data frames with dPET data frames and storing a co-registered dPET volume of frames in the computer memory (para. 0043; process 500 for performing a dynamic PET MR co-registration operation. In step 501, the dynamic PET engine is configured to motion correct the collected dynamic PET data to eliminate motion related noise present in the data. In step 502, a reference transformation is calculated by co-registering the average (averaging across all time frames) motion corrected PET volume with a T1-weighted MRI using non-rigid transform to generate a transformation matrix. In step 503, the dynamic PET engine is configured to apply the reference transformation (e.g., a transformation matrix) across all time frames to generate a co-registered dynamic PET volume in the MR subject space (i.e., the MR of the target site/organ).);
applying the co-registered MRI data frames as inputs to a three-dimensional convolutional neural network (3D-CNN) that outputs localized data frames (paras. 0044, 0062, 0075; In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. Notably, the dynamic PET engine can be configured to provide a fully automated method for correctly identifying the carotid arteries in dynamic PET images, allowing researchers to save time spent on drawing, redrawing, and validating time activity curves. In addition, the model will identify all voxels relevant to the region of interest, making the blood input more robust to noisy areas and more representative of the whole region. Lastly, the robustness of neural networks to random error will allow the neural network to be trained on data drawn manually with slight variations between researchers, but can provide results with high reproducibility. The examiner notes that a 3D CNN is one type of artificial neural network that is used for segmentation and classification. The ANN is used to classify images that contains the carotid arteries correctly from the set of images inputted to the model.);
saving, in the computer memory, visible data frames comprising the localized data frames having the preferred anatomical portion being visible therein (paras. 0044, 0062, 0075; In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. Notably, the dynamic PET engine can be configured to provide a fully automated method for correctly identifying the carotid arteries in dynamic PET images, allowing researchers to save time spent on drawing, redrawing, and validating time activity curves. In addition, the model will identify all voxels relevant to the region of interest, making the blood input more robust to noisy areas and more representative of the whole region. Lastly, the robustness of neural networks to random error will allow the neural network to be trained on data drawn manually with slight variations between researchers, but can provide results with high reproducibility. The examiner notes that a 3D CNN is one type of artificial neural network that is used for segmentation and classification. The ANN is used to classify images that contains the carotid arteries correctly from the set of images inputted to the model.);
segmenting the preferred anatomical portions from the visible data frames and saving segmented data frames in the computer memory (para. 0037; trained ANN model 210 may reside on memory of computing platform(s) 202 and be executable by processor(s) 204. Trained ANN model 210 may be configured to execute an automated segmentation method (e.g., segment out internal carotid arteries from dynamic PET data) and to identify epileptic seizure foci at the brain lobe level in the manner described below.);
calculating an image derived input function (IDIF) for blood flow into the preferred anatomical portions (para. 0044; step 306 indicates that a dynamic PET engine is also configured to generate objective parametric PET maps from a model corrected blood input function (MCIF). Notably, the dynamic PET engine is configured to initially generate an image-derived blood input function (IDIF) from which an MCIF can be computed. FIG. 6 illustrates a diagram depicting an example process 600 for generating an IDIF. In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. In step 603, the dynamic PET engine is configured to draw a blood sampling region and tissue. In step 604, the dynamic PET engine is configured to compute an average of multiple samples from different PET frame slices to generate a plurality of blood time activity curves of a model IDIF. Optimization of the IDIF's model equations/curves can be used by the dynamic PET engine to yield an estimation of the MCIF (as shown in FIG. 7 and described in greater detail below).);
calculating a model-corrected input function (MCIF) for blood flow using the IDIF (para. 0044; step 306 indicates that a dynamic PET engine is also configured to generate objective parametric PET maps from a model corrected blood input function (MCIF). Notably, the dynamic PET engine is configured to initially generate an image-derived blood input function (IDIF) from which an MCIF can be computed. FIG. 6 illustrates a diagram depicting an example process 600 for generating an IDIF. In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. In step 603, the dynamic PET engine is configured to draw a blood sampling region and tissue. In step 604, the dynamic PET engine is configured to compute an average of multiple samples from different PET frame slices to generate a plurality of blood time activity curves of a model IDIF. Optimization of the IDIF's model equations/curves can be used by the dynamic PET engine to yield an estimation of the MCIF (as shown in FIG. 7 and described in greater detail below).);
computing a Ki map for the preferred anatomical portions, wherein the Ki map illustrates influx of the tracer into the preferred anatomical portion (para. 0048; each voxel of dynamic PET data along with blood input is independently fed into a graphical Patlak model to compute parametric K.sub.i maps using linear regression and subsequently z score maps. Specifically, whole brain voxel-level rate of FDG uptake (K.sub.i) can be computed from the motion-corrected, co-registered dFDG PET data for each patient subject. K.sub.i maps for each patient can be converted (e.g., by the dynamic PET engine) to voxel level z-score maps by normalizing to the whole brain mean and standard deviation within each patient subject.).
however, Quigg fails to explicitly teach a probability distribution for respective localized data frames, wherein the probability distribution corresponds to a preferred anatomical portion being visible in each of the respective localized data frames.
Feng, in the same field of endeavor, teach a probability distribution for respective localized data frames, wherein the probability distribution corresponds to a preferred anatomical portion being visible in each of the respective localized data frames (para. 0187; two-class activation refers to mapping the input sample data to a value between 0 and 1 through a two-class activation function, indicating the probability that the sample belongs to the positive class. Among them, in the embodiment of the present invention, two-category activation is performed on the vector representation result to classify the target object picture into the first category, that is, the positive category, or into the second category, that is, the negative category. Correspondingly, in the embodiment of the present invention, the first category is usually marked as 1 and the second category is marked as 0. Exemplarily, in a one-two classification activation application scenario, after converting each first feature map into a corresponding vector representation result, the embodiment of the present invention inputs these vector representation results into a two-class softmax activation function for classification. , to map the corresponding vector representation result to a probability distribution in the range of 0 to 1, thereby obtaining the corresponding classification result, and dividing the target object picture into the first category, that is, the positive category (such as indicating good appearance), or the second category, that is, negative category (for example, it means that the appearance is not good-looking).
It would have been obvious to an ordinary skilled in the art before the invention was made to modify the artificial neural network model of Quigg to incorporate the teaching of Feng to include an artificial neural network model comprising a softmax function to generate a probability distribution. Doing so would improve target frame selection accuracy and quality as disclosed within Feng in para. 0194.
Regarding claim 2, Quigg teaches the computer implemented method of Claim 1, wherein the selected anatomy comprises a brain of the subject and the preferred anatomical portion comprises an internal carotid artery of the subject (paras. 0042 and 0044; For example, MRI image data can be co-registered with a high-resolution T1-weighted MRI template (e.g., a template provided by the Montreal Neurological Institute (MNI) using a non-rigid transform, and a second transformation matrix can be generated (which can be used to bring the masks of a Destrieux atlas defined on the MR brain template from the MNI to the subject's MR space). For example, all of the 164 regions of a Destrieux atlas, which is defined on the same MR brain template, may be binned to generate 36 regions of interest (e.g., 18 regions/side). In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood.).
Regarding claim 3, Quigg teaches the computer implemented method of Claim 1, wherein the MRI data comprises three dimensional (3D) Magnetization Prepared Rapid Gradient Echo (MP-RAGE) comprising 3D T1 images of the selected anatomy, and the dPET data comprises four dimensional, fluorodeoxyglucose, positron emission tomography (dFDG-PET 4D) (paras. 0036 and 0038; For example, PET scanner device 220 may include a Siemens Biograph time of flight (TOF) mCT scanner that can be utilized to perform dynamic acquisitions of a target site/organ, wherein the subject may be administered with an intravenous ˜10 mCi radiotracer injection over 10 seconds with an initiation of a 60-minute scan in list-mode format. Further, MRI scanner device 222 may include a Siemens 3T scanner that is configured to captures a high resolution post-contrast T1-weighted MPRAGE MR images (256 pixels×256 pixels×192 slices). dynamic PET engine 208 can be embodied as a software program, process, and/or algorithm that is configured to execute or manage the disclosed dynamic FDG-PET (dFDG-PET)).
Regarding claim 4, Quigg teaches the computer implemented method of Claim 3, wherein the MP-RAGE data comprises previously stored static MRI scans (fig. 3, para. 0041; with the MRI image (e.g., see MRI data acquired in step 302 in FIG. 3) of the target site/organ (e.g., MRI data generated by MRI scanner device 224). The examiner notes that the processor receives stored PET and MRI images for co-registration).
Regarding claim 5, Quigg teaches the computer implemented method of Claim 1, wherein the tracer comprises fluorodeoxyglucose (para. 0033; the disclosed dynamic FDG-PET (e.g., dynamic 2-[18F] fluoro-2-deoxy-D-glucose positron emission tomography, or dFDG-PET) measures volumetric radioactivity on a continuous basis).
Regarding claim 6, Quigg teaches the computer implemented method of Claim 1, wherein co-registering the MRI data and the dPET data comprises aligning the MRI data and the dPET data to a template and labeling aligned data to an atlas to identify the preferred anatomical portion (paras. 0041 and 0042; The average of all the motion corrected PET frames were also resliced and subsequently co-registered (e.g., see step 403) with the MRI image (e.g., see MRI data acquired in step 302 in FIG. 3) of the target site/organ (e.g., MRI data generated by MRI scanner device 224). In some embodiments, the resliced PET frames may be co-registered with a T1-weighted MRI using non-rigid transform to generate a transformation matrix (which may be utilized to co-register the subject's PET data with a subject's MR image). This transformation matrix is used, in turn, by dynamic PET engine 208 to generate a co-registered dynamic PET volume (and/or dynamic PET data). dynamic PET engine 208 is further configured to generate an atlas and template. For example, MRI image data can be co-registered with a high-resolution T1-weighted MRI template (e.g., a template provided by the Montreal Neurological Institute (MNI) using a non-rigid transform, and a second transformation matrix can be generated (which can be used to bring the masks of a Destrieux atlas defined on the MR brain template from the MNI to the subject's MR space). For example, all of the 164 regions of a Destrieux atlas, which is defined on the same MR brain template, may be binned to generate 36 regions of interest (e.g., 18 regions/side). The second transformation matrix can be inverted (e.g., by dynamic PET engine 208) and applied to all ROIs to transfer them from the standard MNI template into the patient MRI (e.g., step 305 in FIG. 3). The ROIs can then be applied by dynamic PET engine 208 onto the parametric maps (e.g., voxel-by-voxel maps) generated from dFDG-PET images during the co-registration stage shown in step 305 in FIG. 3).
Regarding claim 7, Quigg teaches the computer implementation method of Claim 6, wherein the aligning is in MRI space (paras. 0041 and 0042; The average of all the motion corrected PET frames were also resliced and subsequently co-registered (e.g., see step 403) with the MRI image (e.g., see MRI data acquired in step 302 in FIG. 3) of the target site/organ (e.g., MRI data generated by MRI scanner device 224). In some embodiments, the resliced PET frames may be co-registered with a T1-weighted MRI using non-rigid transform to generate a transformation matrix (which may be utilized to co-register the subject's PET data with a subject's MR image). This transformation matrix is used, in turn, by dynamic PET engine 208 to generate a co-registered dynamic PET volume (and/or dynamic PET data). dynamic PET engine 208 is further configured to generate an atlas and template. For example, MRI image data can be co-registered with a high-resolution T1-weighted MRI template (e.g., a template provided by the Montreal Neurological Institute (MNI) using a non-rigid transform, and a second transformation matrix can be generated (which can be used to bring the masks of a Destrieux atlas defined on the MR brain template from the MNI to the subject's MR space). For example, all of the 164 regions of a Destrieux atlas, which is defined on the same MR brain template, may be binned to generate 36 regions of interest (e.g., 18 regions/side). The second transformation matrix can be inverted (e.g., by dynamic PET engine 208) and applied to all ROIs to transfer them from the standard MNI template into the patient MRI (e.g., step 305 in FIG. 3). The ROIs can then be applied by dynamic PET engine 208 onto the parametric maps (e.g., voxel-by-voxel maps) generated from dFDG-PET images during the co-registration stage shown in step 305 in FIG. 3).
Regarding claim 8, Quigg teaches the computer implemented method of Claim 1, wherein the 3D-CNN comprises a neural network classifier (para. 0068; 4) Development of new feed forward networks specific to classification of dynamic voxels. This application of a deep feed forward network may be optimized through testing a matrix of activation functions, learning rates, layer dimensions, and error reduction techniques. This network can also help to overcome the challenge of learning an imbalanced dataset using resampling techniques.).
Regarding claim 9, Quigg teaches the computer implemented method of Claim 1, further comprising segmenting the visible data frames with a UNETR neural network, and using segmented visible data frames to calculate IDIF (para. 0037; trained ANN model 210 may reside on memory of computing platform(s) 202 and be executable by processor(s) 204. Trained ANN model 210 may be configured to execute an automated segmentation method (e.g., segment out internal carotid arteries from dynamic PET data) and to identify epileptic seizure foci at the brain lobe level in the manner described below.).
Regarding claim 10, Quigg teaches the computer implemented method of Claim 1, further comprising applying a Recurrent Neural Network (RNN) to the IDIF to derive the MCIF with partial volume corrections (paras. 0044 and 0051; the calculation of tracer kinetics can be performed by the dynamic PET engine and requires the capturing of a reference value of blood tracer concentration, the blood input function (IF). The blood IF describes the level of radiotracer within the subject's blood that is available for tissues to use, as a means of calibrating regional dynamic glucose changes. More specifically, tracer uptake can be calibrated with a model corrected blood input function with partial volume corrections to generate tracer parametric maps compared between mean radiation values between hemispheres with z-scores. ).
Regarding claim 11, Quigg teaches the computer implemented method of Claim 1, however, fails to explicitly teach further comprising calculating the probability distribution for respective localized data frames with a softmax function.
Feng, in the same field of endeavor, teach calculating the probability distribution for respective localized data frames with a softmax function. (para. 0187; two-class activation refers to mapping the input sample data to a value between 0 and 1 through a two-class activation function, indicating the probability that the sample belongs to the positive class. Among them, in the embodiment of the present invention, two-category activation is performed on the vector representation result to classify the target object picture into the first category, that is, the positive category, or into the second category, that is, the negative category. Correspondingly, in the embodiment of the present invention, the first category is usually marked as 1 and the second category is marked as 0. Exemplarily, in a one-two classification activation application scenario, after converting each first feature map into a corresponding vector representation result, the embodiment of the present invention inputs these vector representation results into a two-class softmax activation function for classification. , to map the corresponding vector representation result to a probability distribution in the range of 0 to 1, thereby obtaining the corresponding classification result, and dividing the target object picture into the first category, that is, the positive category (such as indicating good appearance), or the second category, that is, negative category (for example, it means that the appearance is not good-looking).
It would have been obvious to an ordinary skilled in the art before the invention was made to modify the artificial neural network model of Quigg to incorporate the teaching of Feng to include an artificial neural network model comprising a softmax function to generate a probability distribution. Doing so would improve target frame selection accuracy and quality as disclosed within Feng in para. 0194.
Regarding claim 12, Quigg teaches the computer implemented method of Claim 11, however, fails to explicitly teach further comprising setting a threshold for the probability distribution, above which a respective localized data frame qualifies as a visible data frame.
Feng, in the same field of endeavor, teach setting a threshold for the probability distribution, above which a respective localized data frame qualifies as a visible data frame (para. 0194; Among them, in the embodiment of the present invention, the average image quality index can be used to evaluate the clarity, contrast, noise or appearance of the target object in the image. Next, the embodiment of the present invention uses the average image quality index as a scoring threshold for target video frame selection. For example, a video frame whose image quality score is greater than the scoring threshold among candidate video frames is selected as the target video frame, thereby improving target video frame selection accuracy and quality.).
It would have been obvious to an ordinary skilled in the art before the invention was made to modify the artificial neural network model of Quigg to incorporate the teaching of Feng to include an artificial neural network model comprising a softmax function to generate a probability distribution and use threshold to identify the optimal frame. Doing so would improve target frame selection accuracy and quality as disclosed within Feng in para. 0194.
Regarding claim 13, Quigg teaches a system of using digital images for mapping metabolic activity within a selected anatomy of a subject, the system comprising (paras. 0029 and 0034; dynamic PET processing can be conducted using one or more host computing devices. For example, FIG. 2 is a block diagram of an example computer platform system 200 for performing dynamic PET and associated processes. the disclosed subject matter relates to an alternative process of metabolic mapping, dynamic PET. This novel technology quantifies the dynamics of radiotracer uptake and decay in order to provide additional specificity or sensitivity in mapping of the desired metabolic activity in the target site (e.g., target organ).):
a computer having a processor connected to computer memory storing software that, when executed, performs computer instruction steps of a machine learning architecture comprising (para. 0035; system 200 may include one or more computing platform(s) 202 (e.g., a dynamic PET platform) having one or more processor(s) 204, such as a central processing unit (e.g., a single core or multiple processing cores), a microprocessor, a microcontroller, a network processor, an application-specific integrated circuit (ASIC), or the like. Computing platform 202 may also include memory 206. Memory 206 may comprise random access memory (RAM), flash memory, a magnetic disk storage drive, and the like. In some embodiments, memory 206 may be configured to store a dynamic PET engine 208 and a trained artificial neural network (ANN) model 210 (e.g., a dynamic PET ANN).):
collecting magnetic resonance image (MRI) data of the selected anatomy of the subject including a preferred anatomical portion of the selected anatomy (para. 0036; dynamic PET engine 208 may be configured to receive image data from each of a PET scanner device 220 and/or a MRI scanner device 222.);
collecting dynamic positron emission tomography (dPET) data of the selected anatomy for the subject, in the time domain, with a tracer applied to the selected anatomy of the subject (para. 0036; dynamic PET engine 208 may be configured to receive image data from each of a PET scanner device 220 and/or a MRI scanner device 222.);
co-registering MRI data frames with dPET data frames and storing a co-registered dPET volume of frames in the computer memory (para. 0043; process 500 for performing a dynamic PET MR co-registration operation. In step 501, the dynamic PET engine is configured to motion correct the collected dynamic PET data to eliminate motion related noise present in the data. In step 502, a reference transformation is calculated by co-registering the average (averaging across all time frames) motion corrected PET volume with a T1-weighted MRI using non-rigid transform to generate a transformation matrix. In step 503, the dynamic PET engine is configured to apply the reference transformation (e.g., a transformation matrix) across all time frames to generate a co-registered dynamic PET volume in the MR subject space (i.e., the MR of the target site/organ).);
applying the co-registered MRI data frames as inputs to a three-dimensional convolutional neural network (3D-CNN) that outputs localized data frames (paras. 0044, 0062, 0075; In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. Notably, the dynamic PET engine can be configured to provide a fully automated method for correctly identifying the carotid arteries in dynamic PET images, allowing researchers to save time spent on drawing, redrawing, and validating time activity curves. In addition, the model will identify all voxels relevant to the region of interest, making the blood input more robust to noisy areas and more representative of the whole region. Lastly, the robustness of neural networks to random error will allow the neural network to be trained on data drawn manually with slight variations between researchers, but can provide results with high reproducibility. The examiner notes that a 3D CNN is one type of artificial neural network that is used for segmentation and classification. The ANN is used to classify images that contains the carotid arteries correctly from the set of images inputted to the model.);
saving, in the computer memory, visible data frames comprising the localized data frames having the preferred anatomical portion being visible therein (paras. 0044, 0062, 0075; In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. Notably, the dynamic PET engine can be configured to provide a fully automated method for correctly identifying the carotid arteries in dynamic PET images, allowing researchers to save time spent on drawing, redrawing, and validating time activity curves. In addition, the model will identify all voxels relevant to the region of interest, making the blood input more robust to noisy areas and more representative of the whole region. Lastly, the robustness of neural networks to random error will allow the neural network to be trained on data drawn manually with slight variations between researchers, but can provide results with high reproducibility. The examiner notes that a 3D CNN is one type of artificial neural network that is used for segmentation and classification. The ANN is used to classify images that contains the carotid arteries correctly from the set of images inputted to the model.);
segmenting the preferred anatomical portions from the visible data frames and saving segmented data frames in the computer memory (para. 0037; trained ANN model 210 may reside on memory of computing platform(s) 202 and be executable by processor(s) 204. Trained ANN model 210 may be configured to execute an automated segmentation method (e.g., segment out internal carotid arteries from dynamic PET data) and to identify epileptic seizure foci at the brain lobe level in the manner described below.);
calculating an image derived input function (IDIF) for blood flow into the preferred anatomical portions (para. 0044; step 306 indicates that a dynamic PET engine is also configured to generate objective parametric PET maps from a model corrected blood input function (MCIF). Notably, the dynamic PET engine is configured to initially generate an image-derived blood input function (IDIF) from which an MCIF can be computed. FIG. 6 illustrates a diagram depicting an example process 600 for generating an IDIF. In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. In step 603, the dynamic PET engine is configured to draw a blood sampling region and tissue. In step 604, the dynamic PET engine is configured to compute an average of multiple samples from different PET frame slices to generate a plurality of blood time activity curves of a model IDIF. Optimization of the IDIF's model equations/curves can be used by the dynamic PET engine to yield an estimation of the MCIF (as shown in FIG. 7 and described in greater detail below).);
calculating a model-corrected input function (MCIF) for blood flow using the IDIF (para. 0044; step 306 indicates that a dynamic PET engine is also configured to generate objective parametric PET maps from a model corrected blood input function (MCIF). Notably, the dynamic PET engine is configured to initially generate an image-derived blood input function (IDIF) from which an MCIF can be computed. FIG. 6 illustrates a diagram depicting an example process 600 for generating an IDIF. In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. In step 603, the dynamic PET engine is configured to draw a blood sampling region and tissue. In step 604, the dynamic PET engine is configured to compute an average of multiple samples from different PET frame slices to generate a plurality of blood time activity curves of a model IDIF. Optimization of the IDIF's model equations/curves can be used by the dynamic PET engine to yield an estimation of the MCIF (as shown in FIG. 7 and described in greater detail below).);
computing a Ki map for the preferred anatomical portions, wherein the Ki map illustrates influx of the tracer into the preferred anatomical portion (para. 0048; each voxel of dynamic PET data along with blood input is independently fed into a graphical Patlak model to compute parametric K.sub.i maps using linear regression and subsequently z score maps. Specifically, whole brain voxel-level rate of FDG uptake (K.sub.i) can be computed from the motion-corrected, co-registered dFDG PET data for each patient subject. K.sub.i maps for each patient can be converted (e.g., by the dynamic PET engine) to voxel level z-score maps by normalizing to the whole brain mean and standard deviation within each patient subject.).
however, Quigg fails to explicitly teach a probability distribution for respective localized data frames, wherein the probability distribution corresponds to a preferred anatomical portion being visible in each of the respective localized data frames.
Feng, in the same field of endeavor, teach a probability distribution for respective localized data frames, wherein the probability distribution corresponds to a preferred anatomical portion being visible in each of the respective localized data frames (para. 0187; two-class activation refers to mapping the input sample data to a value between 0 and 1 through a two-class activation function, indicating the probability that the sample belongs to the positive class. Among them, in the embodiment of the present invention, two-category activation is performed on the vector representation result to classify the target object picture into the first category, that is, the positive category, or into the second category, that is, the negative category. Correspondingly, in the embodiment of the present invention, the first category is usually marked as 1 and the second category is marked as 0. Exemplarily, in a one-two classification activation application scenario, after converting each first feature map into a corresponding vector representation result, the embodiment of the present invention inputs these vector representation results into a two-class softmax activation function for classification. , to map the corresponding vector representation result to a probability distribution in the range of 0 to 1, thereby obtaining the corresponding classification result, and dividing the target object picture into the first category, that is, the positive category (such as indicating good appearance), or the second category, that is, negative category (for example, it means that the appearance is not good-looking).
It would have been obvious to an ordinary skilled in the art before the invention was made to modify the artificial neural network model of Quigg to incorporate the teaching of Feng to include an artificial neural network model comprising a softmax function to generate a probability distribution. Doing so would improve target frame selection accuracy and quality as disclosed within Feng in para. 0194.
Regarding claim 14, Quigg teaches the system of Claim 13, further comprising a PET scanner and an MRI scanner in communication with the computer (fig. 2, paras. 0035-0036; system 200 may include one or more computing platform(s) 202 (e.g., a dynamic PET platform) having one or more processor(s) 204, such as a central processing unit (e.g., a single core or multiple processing cores), a microprocessor, a microcontroller, a network processor, an application-specific integrated circuit (ASIC), or the like. Computing platform 202 may also include memory 206. Memory 206 may comprise random access memory (RAM), flash memory, a magnetic disk storage drive, and the like. In some embodiments, memory 206 may be configured to store a dynamic PET engine 208 and a trained artificial neural network (ANN) model 210 (e.g., a dynamic PET ANN). Dynamic PET engine 208 may include one or more algorithms, software programs, software processes, and the like.).
Regarding claim 15, Quigg teaches the system of Claim 14, wherein the MRI scanner is configured for scanning T1 images of the subject (para. 0036; Further, MRI scanner device 222 may include a Siemens 3T scanner that is configured to captures a high resolution post-contrast T1-weighted MPRAGE MR images (256 pixels×256 pixels×192 slices).).
Regarding claim 16, Quigg teaches a computer program product comprising: a non-transitory computer readable medium storing software that, when executed, performs computer instruction steps of a machine learning architecture comprising (para. 0035; system 200 may include one or more computing platform(s) 202 (e.g., a dynamic PET platform) having one or more processor(s) 204, such as a central processing unit (e.g., a single core or multiple processing cores), a microprocessor, a microcontroller, a network processor, an application-specific integrated circuit (ASIC), or the like. Computing platform 202 may also include memory 206. Memory 206 may comprise random access memory (RAM), flash memory, a magnetic disk storage drive, and the like. In some embodiments, memory 206 may be configured to store a dynamic PET engine 208 and a trained artificial neural network (ANN) model 210 (e.g., a dynamic PET ANN). Dynamic PET engine 208 may include one or more algorithms, software programs, software processes, and the like. As described below, dynamic PET engine 208 is configured to control, manage, and administer a plurality of processes corresponding to the execution of the disclosed dynamic PET methodology and functionality.):
collecting magnetic resonance image (MRI) data of the selected anatomy of the subject including a preferred anatomical portion of the selected anatomy (para. 0036; dynamic PET engine 208 may be configured to receive image data from each of a PET scanner device 220 and/or a MRI scanner device 222.);
collecting dynamic positron emission tomography (dPET) data of the selected anatomy for the subject, in the time domain, with a tracer applied to the selected anatomy of the subject (para. 0036; dynamic PET engine 208 may be configured to receive image data from each of a PET scanner device 220 and/or a MRI scanner device 222.);
co-registering MRI data frames with dPET data frames and storing a co-registered dPET volume of frames in the computer memory (para. 0043; process 500 for performing a dynamic PET MR co-registration operation. In step 501, the dynamic PET engine is configured to motion correct the collected dynamic PET data to eliminate motion related noise present in the data. In step 502, a reference transformation is calculated by co-registering the average (averaging across all time frames) motion corrected PET volume with a T1-weighted MRI using non-rigid transform to generate a transformation matrix. In step 503, the dynamic PET engine is configured to apply the reference transformation (e.g., a transformation matrix) across all time frames to generate a co-registered dynamic PET volume in the MR subject space (i.e., the MR of the target site/organ).);
applying the co-registered MRI data frames as inputs to a three-dimensional convolutional neural network (3D-CNN) that outputs localized data frames (paras. 0044, 0062, 0075; In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. Notably, the dynamic PET engine can be configured to provide a fully automated method for correctly identifying the carotid arteries in dynamic PET images, allowing researchers to save time spent on drawing, redrawing, and validating time activity curves. In addition, the model will identify all voxels relevant to the region of interest, making the blood input more robust to noisy areas and more representative of the whole region. Lastly, the robustness of neural networks to random error will allow the neural network to be trained on data drawn manually with slight variations between researchers, but can provide results with high reproducibility. The examiner notes that a 3D CNN is one type of artificial neural network that is used for segmentation and classification. The ANN is used to classify images that contains the carotid arteries correctly from the set of images inputted to the model.);
saving, in the computer memory, visible data frames comprising the localized data frames having the preferred anatomical portion being visible therein (paras. 0044, 0062, 0075; In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. Notably, the dynamic PET engine can be configured to provide a fully automated method for correctly identifying the carotid arteries in dynamic PET images, allowing researchers to save time spent on drawing, redrawing, and validating time activity curves. In addition, the model will identify all voxels relevant to the region of interest, making the blood input more robust to noisy areas and more representative of the whole region. Lastly, the robustness of neural networks to random error will allow the neural network to be trained on data drawn manually with slight variations between researchers, but can provide results with high reproducibility. The examiner notes that a 3D CNN is one type of artificial neural network that is used for segmentation and classification. The ANN is used to classify images that contains the carotid arteries correctly from the set of images inputted to the model.);
segmenting the preferred anatomical portions from the visible data frames and saving segmented data frames in the computer memory (para. 0037; trained ANN model 210 may reside on memory of computing platform(s) 202 and be executable by processor(s) 204. Trained ANN model 210 may be configured to execute an automated segmentation method (e.g., segment out internal carotid arteries from dynamic PET data) and to identify epileptic seizure foci at the brain lobe level in the manner described below.);
calculating an image derived input function (IDIF) for blood flow into the preferred anatomical portions (para. 0044; step 306 indicates that a dynamic PET engine is also configured to generate objective parametric PET maps from a model corrected blood input function (MCIF). Notably, the dynamic PET engine is configured to initially generate an image-derived blood input function (IDIF) from which an MCIF can be computed. FIG. 6 illustrates a diagram depicting an example process 600 for generating an IDIF. In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. In step 603, the dynamic PET engine is configured to draw a blood sampling region and tissue. In step 604, the dynamic PET engine is configured to compute an average of multiple samples from different PET frame slices to generate a plurality of blood time activity curves of a model IDIF. Optimization of the IDIF's model equations/curves can be used by the dynamic PET engine to yield an estimation of the MCIF (as shown in FIG. 7 and described in greater detail below).);
calculating a model-corrected input function (MCIF) for blood flow using the IDIF (para. 0044; step 306 indicates that a dynamic PET engine is also configured to generate objective parametric PET maps from a model corrected blood input function (MCIF). Notably, the dynamic PET engine is configured to initially generate an image-derived blood input function (IDIF) from which an MCIF can be computed. FIG. 6 illustrates a diagram depicting an example process 600 for generating an IDIF. In step 601, the dynamic PET engine is configured to read the co-registered PET data into matrices. In step 602, the internal carotid arteries are identified at the early time frames (within 16-20 seconds of tracer administration) to capture the wash-in and wash-out kinetics of the tracer from the blood. In step 603, the dynamic PET engine is configured to draw a blood sampling region and tissue. In step 604, the dynamic PET engine is configured to compute an average of multiple samples from different PET frame slices to generate a plurality of blood time activity curves of a model IDIF. Optimization of the IDIF's model equations/curves can be used by the dynamic PET engine to yield an estimation of the MCIF (as shown in FIG. 7 and described in greater detail below).);
computing a Ki map for the preferred anatomical portions, wherein the Ki map illustrates influx of the tracer into the preferred anatomical portion (para. 0048; each voxel of dynamic PET data along with blood input is independently fed into a graphical Patlak model to compute parametric K.sub.i maps using linear regression and subsequently z score maps. Specifically, whole brain voxel-level rate of FDG uptake (K.sub.i) can be computed from the motion-corrected, co-registered dFDG PET data for each patient subject. K.sub.i maps for each patient can be converted (e.g., by the dynamic PET engine) to voxel level z-score maps by normalizing to the whole brain mean and standard deviation within each patient subject.).
however, Quigg fails to explicitly teach a probability distribution for respective localized data frames, wherein the probability distribution corresponds to a preferred anatomical portion being visible in each of the respective localized data frames.
Feng, in the same field of endeavor, teach a probability distribution for respective localized data frames, wherein the probability distribution corresponds to a preferred anatomical portion being visible in each of the respective localized data frames (para. 0187; two-class activation refers to mapping the input sample data to a value between 0 and 1 through a two-class activation function, indicating the probability that the sample belongs to the positive class. Among them, in the embodiment of the present invention, two-category activation is performed on the vector representation result to classify the target object picture into the first category, that is, the positive category, or into the second category, that is, the negative category. Correspondingly, in the embodiment of the present invention, the first category is usually marked as 1 and the second category is marked as 0. Exemplarily, in a one-two classification activation application scenario, after converting each first feature map into a corresponding vector representation result, the embodiment of the present invention inputs these vector representation results into a two-class softmax activation function for classification. , to map the corresponding vector representation result to a probability distribution in the range of 0 to 1, thereby obtaining the corresponding classification result, and dividing the target object picture into the first category, that is, the positive category (such as indicating good appearance), or the second category, that is, negative category (for example, it means that the appearance is not good-looking).
It would have been obvious to an ordinary skilled in the art before the invention was made to modify the artificial neural network model of Quigg to incorporate the teaching of Feng to include an artificial neural network model comprising a softmax function to generate a probability distribution. Doing so would improve target frame selection accuracy and quality as disclosed within Feng in para. 0194.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ZAINAB M ALDARRAJI whose telephone number is (571)272-8726. The examiner can normally be reached Monday-Thursday7AM-5PM 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, Carey Michael can be reached at (571) 270-7235. 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.
/ZAINAB MOHAMMED ALDARRAJI/ Patent Examiner, Art Unit 3797
Prosecution Insights