DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
Applicant’s submission filed 11/03/2025 has been entered. The claims 1-20 are pending in the current application.
Response to Arguments
Applicant's arguments filed 11/03/2025 have been fully considered but they are not persuasive.
In Remarks, applicant argued in essence with the claim interpretation with respect to the specific exemplary embodiment of the uptake time in the Background Section of the Specification as the time between tracer injection and image acquisition as Background information. Nowhere elsewhere in applicant’s Specification defines the claimed uptake time in the same manner as the Background Section. Applicant’s claim invention set forth in the claim 1 fails to define the claimed uptake time.
An image with improved image quality is associated with an extended uptake time.
This is because.
Optimizing Image Quality: Extending the uptake time to around 90 minutes can improve the target-to-background ratio (the contrast between tumor and healthy tissue), particularly in liver lesions.
Lesion Detection: While 60 minutes is standard, some studies suggest that longer, or "delayed," imaging can help distinguish cancer from benign inflammation, as tumor tracer uptake often continues to increase over time while background uptake decreases.
However, the claim limitation is subject to broadest reasonable interpretation consistent with applicant’s specification. Limitation from the specification cannot be imported into the claims.
Since applicant unnecessarily import additional claim limitation from the Specification with respect to the claimed scanned image. Applicant should refer to the Background prior art references for the multiple definitions of uptake time. The uptake time interval can be defined to be the same as the acquisition time interval. This is the basis in the examiner’s ground(s) of rejection in the Non-Final Office Action dated 8/13/2025.
Moreover, applicant’s scanned images are not necessarily PET/SPECT images, and could be scanned images of any image acquisition modalities. The image modalities include an image modality (e.g., MR acquisition modality) with the uptake time interval equal to the acquisition time interval. The claim limitation is thus subject to broadest reasonable interpretation consistent with applicant’s specification and limitations from the specification cannot be imported into the claims.
Uptake time is the same as imaging acquisition time as Background Information.
For example, in von Schulthess et al. US-PGPUB No. 2009/0264753 discloses at Paragraph 0008 that a first imaging modality during a first acquisition time interval occurs proximate in time with at least one of the uptake time interval.
Schulthess teaches at Paragraph 0009 that the first acquisition time interval occurs coincident in time with the uptake time interval such that the first image data set reflects the physiologic state of the ROI during the uptake time interval.
Schulthess teaches at Paragraph 0040 that the MR acquisition time interval 410 ends coincident with the end of the uptake time interval 406.
Kaplan et al. US-PGPUB No. 2021/0052233 teaches at Paragraph 0032 that during this waiting period, the CT gantry 4 may optionally be employed to acquire scout CT images.
Ye et al. US-PGPUB No. 2024/0156426 teaches at Paragraph 0011 that designating an acquisition time associated with the target reconstruction image as the injection time of the drug.
Accordingly, the examiner interpreted the claimed uptake time as the image acquisition time and the rejection is maintained.
Even if the claim limitation of uptake time were to be interpreted according to applicant arguments,
Relationship between SPECT image intensity and uptake time.
Moreover, Shi et al. US-PGPUB No. 2022/0207791 teaches at Paragraph 0054 that the SPECT image intensity represents the tracer activity and thus varies among patients due to multiple factors, including different tracer injection dose, time delay from injection to imaging (uptake time).
As a result, the SPECT image intensity (contrast) is a function of uptake time or the uptake time is an inverse function of the SPECT image intensity.
Relationship between SUV value and uptake time.
Moreover, Goedicke et al. US-PGPUB No. 20210398329 teaches at Paragraph 0047 that the SPECT image intensity I is converted to a corresponding SUV value defined by Equation (1) which is proportional to 2t where t is the wait time between administration of the radio pharmaceutical and the PET imaging data acquisition (uptake time). The larger the uptake time, the larger the SUV value.
The SUV value is proportional to the uptake time.
Siddu teaches at Paragraph 0090 that SUV values of the PET images will vary with respect to uptake time of the nuclear tracer.
An image with improved image quality is associated with an extended uptake time.
Sibille teaches at Paragraph 0036-0037 that new SUV value (of the reconstructed SPECT image) is modified (thus corresponding uptake time is modified) from the standard SUV value based on the uptake time data to provide more accurate SUV measures based on the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image using standard acquisition time as reference where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Sibille teaches at Paragraph 0010 that methods and systems of the present disclosure may provide accelerated SPECT image acquisition (low uptake time) while preserving accuracy in standardized uptake quantification (standard high uptake time). The quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contrast (which is a function of the uptake time and thus boosting the image contrast of the fast scan SPECT image affects the uptake time) and thus and the accuracy (e.g., standardized uptake value (SUV) in lesion regions).
An image with improved image quality is associated with an extended uptake time. The image with the boosted image contrast is associated with an extended uptake time.
SUV is also a function of the uptake time according to Goedicke at Paragraph 0047 and Siddu at Paragraph 0090. Moreover, the boosted SUV of Sibille means that the adjusted uptake time.
Sibille teaches at Paragraph 0010 that the method herein may boost the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention map. Sibille teaches at Paragraph 0036 that the proposed method can effectively suppress the noise texture in the fast-scanned SPECT image and recover clear anatomy structural details, especially SUV value in the lesion regions, which is comparable to SPECT image quality using standard acquisition time.
Sibille teaches at Paragraph [0035] Methods and systems herein may effectively reduce the radiation exposure by shortening the SPECT/CT examination time (e.g., shorten the SPECT scan time) without compromising the quality of the output image. In some embodiments, systems and methods herein may synthesize an enhanced medical image from the fast scan SPECT image and CT image and the enhanced medical image may have an image quality same as a SPECT image acquired with standard acquisition time combined with a corresponding CT image. In some embodiments, the enhanced medical image may have an image quality improved over the fast scan SPECT image in terms of quantification accuracy. For example, the provided method may utilize the associated features between a fast SPECT scan and corresponding CT image to improve anatomy structural boundary and the overall image quality.
[0036] In some embodiments, the provided method may allow for faster SPECT imaging acquisition while preserving quantification accuracy related to physiological or biochemical information. For example, methods and systems of the present disclosure may provide accelerated SPECT image acquisition while preserving accuracy in standardized uptake quantification. The method herein may preserve or improve the quantification accuracy by boosting the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions. In some cases, the image contrast and accuracy in lesion regions may be preserved by penalizing the losses in these regions such as utilizing an attention map. The attention map may comprise an attention feature map or region-of-interest (ROI) attention masks. The attention map may comprise information about the ROI (e.g., lesion attention map) or other attention map that comprises clinically meaningful information. For example, the attention map may comprise information about regions where particular tissues/features are located. For example, the proposed method can effectively suppress the noise texture in the fast-scanned SPECT image and recover clear anatomy structural details, especially SUV value in the lesion regions, which is comparable to SPECT image quality using standard acquisition time.
Sibille teaches at Paragraph 0038 that structural similarity (SSIM) may be calculated for the enhanced and non-enhanced accelerated SPECT scans, where higher image quality is represented by higher PSNR and/or SSIM. Other suitable image quality quantification metrics may also be utilized to quantify the image quality.
Sibille teaches at Paragraph [0040] The fast SPECT image 103 may be acquired with an acquisition time at a selected accelerator factor. For example, the image acquisition time of the fast SPECT image 103 may be ½, ⅓, ¼, ⅕, ⅙, 1/7, ⅛, 1/9 of the standard acquisition time. For example, the fast scan SPECT image 103 may be acquired with ⅛ standard acquisition time or faster. Compared to the standard SPECT 107 (SPECT acquired with standard time such as about 20 min/bed), the fast SPECT image 103 may have a degraded image quality (e.g., greater noise and artifacts, less detail of the radiotracer distribution).
Sibille teaches that the attention feature maps are obtained based on the uptake time.
Sibille teaches at Paragraph 0057 that the attention mask can be obtained based on a SUV threshold in the fast SPECT image (associated with an adjusted uptake time). For instance, a SUV threshold may be selected and the fast SPECT image or input image may be filtered based on the threshold. In some cases, the filtered fast SPECT image may be used as the attention mask. The attention mask may need more accurate boundary enhancement compared to the normal structures and background. Including the attention mask may beneficially preserve the quantification accuracy and/or sharpness of the features (e.g., features of lesion, segment of bone lesion). The attention mask can be obtained using any other suitable methods such as utilizing deep learning techniques. Alternatively, the attention mask may be generated manually by manual selection/segmentation on an input image.
Sibille teaches at Paragraph 0085 that the training datasets for may comprise pairs of standard acquisition (standard uptake time) and shortened acquisition SPECT images (shortened uptake time), CT images and/or attention feature map from the same subject.
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.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al. US-PGPUB No. 2025/0148574 (hereinafter Chen) in view of Xiang et al. US-PGPUB No. 2024/0307018 (hereinafter Xiang); Sibille US-PGPUB No. 2025/0131615 (hereinafter Sibille);
Siddu et al. US-PGPUB No. 2025/0061573 (hereinafter Siddu) and Moriyasu US-PGPUB No. 2021/0019924 (hereinafter Moriyasu).
Re Claim 1:
Chen/Xiang/Sibille teaches a computer-implemented method comprising:
receiving measurement data characterizing a scanned image of a subject (Chen teaches at Paragraph 0034 that the dynamic PET image set contains at least two dynamic PET images and a PET/CT dynamic imaging scanning technology may be adopted to perform imaging scanning on a detected object to obtain the dynamic PET image set.
Sibille teaches at Paragraph 0035 that a SPECT image is acquired with standard acquisition time.
Xiang teaches at Paragraph 0066 that there are two scanning protocols: one standard scan with 20 seconds per frame and one fast scan with 3 seconds per frame.
Xiang teaches at Paragraph 0030-0036 that FIG. 1 shows an example of a fast scan SPECT image acquired with 1/7 standard acquisition time or faster and at Paragraph 0041 that the SPECT image and CT image may be acquired by a SPECT/CT scanner);
receiving uptake time data characterizing a first uptake time of the scanned image (
It is known that the SUV image has an SUV value associated with the initial uptake time.
Chen teaches at Paragraph 0038 that the dynamic PET image corresponding to a preset acquisition time range in the dynamic PET image set or a dynamic SUV image corresponding to the dynamic PET image and at Paragraph 0040 that single acquisition of the dynamic PET image set requires a certain acquisition duration, typically 60 minutes and the preset acquisition time range is used to characterize a preset time period within a total acquisition duration corresponding to a dynamic acquisition image set. Chen teaches at Paragraph 0042 that the dynamic PET image corresponding to this preset acquisition time range 0-5 minutes is an early dynamic PET image in the dynamic PET image set and at Paragraph 0043 the dynamic PET image corresponding to this preset acquisition time range 50-60 minutes is a final dynamic PET image in the dynamic PET image set. Chen teaches at Paragraph 0086 that the input image is a dynamic PET image corresponding a preset acquisition time range (uptake time of 5 minutes) in the dynamic PET image set.
Sibille teaches at Paragraph 0035 that a SPECT image is acquired with standard acquisition time.
Xiang teaches at Paragraph 0066 that there are two scanning protocols: one standard scan with 20 seconds per frame and one fast scan with 3 seconds per frame.
Xiang teaches at Paragraph 0030-0036 that FIG. 1 shows an example of a fast scan SPECT image acquired with 1/7 standard acquisition time or faster….a pair of SPECT and CT images are acquired using a SPECT/CT scanner and fast scan SPECT has low image quality due to the fast scan and at Paragraph 0041 that the SPECT image and CT image may be acquired by a SPECT/CT scanner.
Xiang teaches at Paragraph 0013 that the deep learning network model is trained using training data comprising a SPECT image acquired using shortened acquisition time, a corresponding CT image and a SPECT image acquired using a standard acquisition time.
Xiang teaches at Paragraph 0009 that the term “fast scan SPECT image” may generally refer to a SPECT image acquired under shortened acquisition time at an acceleration factor of a value greater than 1. The standard acquisition time is at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20 times of the shortened acquisition time of the fast scan SPECT image.);
applying a trained machine learning process to the measurement data and the uptake time data and, based on the application of the trained machine learning process to the measurement data and the uptake time data, generating output image data characterizing an output image at a second uptake time (
Chen teaches at Paragraph 0041 that when the original PET parameter image is the K1 parameter image, a maximum acquisition time corresponding to the preset acquisition time range is the total acquisition duration corresponding to the dynamic PET image set and at Paragraph 0055 the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained and the input image is input into the image enhancement model to obtain the output predicted PET parameter image…the image quality of the PET parameter image is improved while preserving image details of the PET parameter image.
Chen teaches at Paragraph 0057 that the original PET parameter image is determined based on an obtained dynamic PET image set and at Paragraph 0086 that the input image is a dynamic PET image corresponding a preset acquisition time range (uptake time of 5 minutes) in the dynamic PET image set and at Paragraph 0087 that the predicted PET parameter image is used as the target PET parameter image corresponding to the original PET parameter image so that the problem that an existing neural network model method requires preparation of a high-quality PET parameter image is solved.
Chen teaches at Paragraph 0088, when the original PET parameter image is a K1 parameter image, the maximum acquisition time corresponding to the preset acquisition time range is (60 minutes) a total acquisition time duration corresponding to the dynamic PET image set and at Paragraph 0099 determining a Euclidean distance difference between the original PET parameter image and the predicted PET parameter image based on the L2 loss function (the predicted PET parameter image has the same up take time as the original PET parameter image which is 60 minutes).
Chen teaches at Paragraph 0081 that at least two parameter feature maps are output based on the inputted input image by at least two encoding convolutional networks in the encoder and the predicted PET parameter image is outputted based on the at least two parameter feature maps
Chen teaches at Paragraph 0047 that the input image is input into an image enhancement model to obtain an output predicted PET parameter image and at Paragraph 0048 that a model architecture of the image enhancement model includes a generative adversarial network architecture, a U-Net architecture, super resolution convolutional neural networks. Chen teaches at Paragraph 0049 that a model parameter of the image enhancement model is adjusted based on the original PET parameter image and predicted PET parameter image.
Chen teaches at Paragraph 0055 that the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained, wherein the input image is the noise image, the dynamic PET image corresponding the preset acquisition time range in the dynamic PET image set or the dynamic SUV image corresponding to the dynamic PET image, the input image is input into the image enhancement model to obtain the output predicted PET parameter image, the model parameter of the image enhancement model is adjusted based on the original PET parameter image and the predicted PET parameter image until the preset number of iterations is met.
Sibille teaches at Paragraph 0010 that methods and systems of the present disclosure may provide accelerated SPECT image acquisition (low uptake time) while preserving accuracy in standardized uptake quantification (standard high uptake time). The quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions). The method herein may boost the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention map. Sibille teaches at Paragraph 0036 that the proposed method can effectively suppress the noise texture in the fast-scanned SPECT image and recover clear anatomy structural details, especially SUV value in the lesion regions, which is comparable to SPECT image quality using standard acquisition time.
An image with improved image quality is associated with an extended uptake time. The image with the boosted image contrast is associated with an extended uptake time.
The boosted SUV value corresponds to the adjusted uptake time.
Sibille teaches at Paragraph 0061 that the fast SPECT image may be acquired by reducing the number of acquisition planes (e.g., standard 120 planes) by a factor of 2, 3, 4, 5, 6, 7, 8 or more.
Sibille teaches at Paragraph 0041 that the deep learning techniques may comprise using a convolutional neural network (CNN) to generate synthesized SPECT images from a fast scan with image quality comparable to SPECT images acquired with standard acquisition time (e.g., standard SPECT). This beneficially accelerates SPECT image acquisition by an acceleration factor of at least 1.5, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, a factor of a value above 20 or below 1.5, or a value between any of the two aforementioned values. The provided method and systems can achieve a shortened acquisition time of no more than ½, ⅓, ¼, ⅕, ⅙, 1/7, ⅛, 1/9, 1/10 of the standard acquisition time.
An image with improved image quality is associated with an extended uptake time. The image with the improved image quality in Xiang is associated with an extended uptake time.
Xiang teaches at Paragraph 0013 that the deep learning network model is trained using training data comprising a SPECT image acquired using shortened acquisition time, a corresponding CT image and a SPECT image acquired using a standard acquisition time.
Xiang teaches at Paragraph 0009 that the term “fast scan SPECT image” may generally refer to a SPECT image acquired under shortened acquisition time at an acceleration factor of a value greater than 1. The standard acquisition time is at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20 times of the shortened acquisition time of the fast scan SPECT image.
Xiang teaches at Paragraph 0031 that the present disclosure provides an imaging acceleration method employing deep learning techniques to improve the image quality acquired with shortened acquisition time, i.e., fast scan and the deep learning techniques may comprise using multi-modality and multi-scale feature aggregation-based framework to generate SPECT images from a fast scan with image quality comparable to SPECT images acquired with standard acquisition time. The methods and systems herein may effectively reduce the radiation exposure by shortening the SPECT/CT examination time, e.g., shorten the SPECT scan time).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have incorporated Xiang/Sibille’s neural network training model to have enhanced the SPECT image with shortened SPECT scan time based on the input SPECT image acquired with fast acquisition time into Chen’s neural network training model to have provided enhanced output SPECT image with enhanced ROI than the ROI of the input SPECT image acquired with fast acquisition time. One of the ordinary skill in the art would have provided an enhanced PET/SPECT image with enhanced ROI region using the trained neural network model.
Chen does not teach the claim limitation:
storing the output image in a data repository.
Moriyasu/Xiang/Sibble teaches the claim limitation of storing the output image in a data repository (Moriyasu teaches at Paragraph 0045 that the storage circuitry 41a is configured to store reconstructed PET image data derived by the PET image reconstructing circuitry 41d.
Sibille teaches at Paragraph 0059 that the training data may be obtained from an imaging system and at Paragraph 0069 that the training data may comprise simulated image sets and the fast acquisition SPECT images may be simulated and reconstructed using standard SPECT tools and at Paragraph 0092 that the one or more databases may utilize any suitable database techniques,. For example, structured query language may be utilized for storing image data, raw collected data and reconstructed image data and training datasets.
Xiang teaches at Paragraph 0056 that the controller may apply a tomographic reconstruction algorithm (e.g., filter backprojection (FBP), iterative algorithm such as algebraic reconstruction technique (ART), etc.) to the multiple projections, yielding a 3-D data set. The SPECT image may be combined with the CT image to generate the combined image as output of the imaging system. For example, reconstruction algorithms have been developed in bone hybrid imaging (SPECT/CT) scanner. Different from classic SPECT reconstructions, such reconstruction algorithms may utilize an ordered subset conjugate gradient minimization algorithm (OSCGM) for image reconstruction. This construction algorithm can provide SPECT images with bone anatomy appearance as a CT-based tissue segmentation is incorporated into SPECT reconstruction and also provide a quantitative reconstruction. Such progress of image acquisition and reconstruction could convey a higher diagnostic confidence through an enhanced bone uptake location.
Xiang teaches at Paragraph 0037 that the provided method may also incorporate a lesion attention loss function to enhance sensitivity of the deep learning model to reconstruct lesion regions with more accurate SUV measures. Xiang teaches at Paragraph 0059 that the database 320 may store image data including raw collected data, reconstructed image data, training datasets).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have stored the output PET/SPECT image in the database resident in a memory. One of the ordinary skill in the art would have been motivated to have stored reconstructed image in a database or displayed in UI which inherently requires the output image to be stored in a display memory buffer.
Siddu teaches at Paragraph 0090 that PET normalization helps to normalize SUV values that will vary with respect to the uptake time of the nuclear tracer and multi-modal fusion of hand-crafted features from CT and PET images such as radiomics features extracted from an ROI bounding box is proposed as well as a machine learning model to map Gleason scores.
Chen teaches at Paragraph 0071 normalizing the input image to obtain a normalized input image and registering the normalized input image with the original PET image to obtain a registered input image. Sibille teaches at Paragraph 0010 that the quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contrast and the accuracy, e.g., standardized uptake value (SUV) (and thus associated with a second uptake time) in lesion regions and the method herein may boost the image contrast and the standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention maps.
According to Siddu, the normalized input image of Chen includes the normalized SUV values as a function of uptake time where Sibille teaches SUV values (features of the enhanced SPECT image) are boosted up to the standard uptake time.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have boosted features of an enhanced PET image as a function of uptake time up to the quality of the PET image acquired at the standard uptake time. One of the ordinary skill in the art would have encoded the features as function of uptake time in a series of encoding convolutional networks.
Re Claim 2:
The claim 2 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that applying the trained machine learning process to the measurement data and the uptake time data comprises: generating feature maps based on the measurement data; and modifying the feature maps based on the uptake time data.
Chen and Sibille/Xiang further teach the claim limitation that applying the trained machine learning process to the measurement data and the uptake time data comprises: generating feature maps based on the measurement data; and modifying the feature maps based on the uptake time data (
Sibille teaches at Paragraph 0037 that new SUV value is modified from the standard SUV value based on the uptake time data to provide more accurate SUV measures based on the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image using standard acquisition time as reference where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Sibille teaches at Paragraph 0010 that methods and systems of the present disclosure may provide accelerated SPECT image acquisition (low uptake time) while preserving accuracy in standardized uptake quantification (standard high uptake time). The quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions). The method herein may boost the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention map. Sibille teaches at Paragraph 0036 that the proposed method can effectively suppress the noise texture in the fast-scanned SPECT image and recover clear anatomy structural details, especially SUV value in the lesion regions, which is comparable to SPECT image quality using standard acquisition time.
Sibille teaches at Paragraph [0035] Methods and systems herein may effectively reduce the radiation exposure by shortening the SPECT/CT examination time (e.g., shorten the SPECT scan time) without compromising the quality of the output image. In some embodiments, systems and methods herein may synthesize an enhanced medical image from the fast scan SPECT image and CT image and the enhanced medical image may have an image quality same as a SPECT image acquired with standard acquisition time combined with a corresponding CT image. In some embodiments, the enhanced medical image may have an image quality improved over the fast scan SPECT image in terms of quantification accuracy. For example, the provided method may utilize the associated features between a fast SPECT scan and corresponding CT image to improve anatomy structural boundary and the overall image quality.
[0036] In some embodiments, the provided method may allow for faster SPECT imaging acquisition while preserving quantification accuracy related to physiological or biochemical information. For example, methods and systems of the present disclosure may provide accelerated SPECT image acquisition while preserving accuracy in standardized uptake quantification. The method herein may preserve or improve the quantification accuracy by boosting the image contrast and the accuracy (e.g., standardized uptake value (SUV) in lesion regions. In some cases, the image contrast and accuracy in lesion regions may be preserved by penalizing the losses in these regions such as utilizing an attention map. The attention map may comprise an attention feature map or region-of-interest (ROI) attention masks. The attention map may comprise information about the ROI (e.g., lesion attention map) or other attention map that comprises clinically meaningful information. For example, the attention map may comprise information about regions where particular tissues/features are located. For example, the proposed method can effectively suppress the noise texture in the fast-scanned SPECT image and recover clear anatomy structural details, especially SUV value in the lesion regions, which is comparable to SPECT image quality using standard acquisition time.
Sibille teaches at Paragraph 0038 that structural similarity (SSIM) may be calculated for the enhanced and non-enhanced accelerated SPECT scans, where higher image quality is represented by higher PSNR and/or SSIM. Other suitable image quality quantification metrics may also be utilized to quantify the image quality.
Sibille teaches at Paragraph [0040] The fast SPECT image 103 may be acquired with an acquisition time at a selected accelerator factor. For example, the image acquisition time of the fast SPECT image 103 may be ½, ⅓, ¼, ⅕, ⅙, 1/7, ⅛, 1/9 of the standard acquisition time. For example, the fast scan SPECT image 103 may be acquired with ⅛ standard acquisition time or faster. Compared to the standard SPECT 107 (SPECT acquired with standard time such as about 20 min/bed), the fast SPECT image 103 may have a degraded image quality (e.g., greater noise and artifacts, less detail of the radiotracer distribution).
Sibille teaches that the attention feature maps are obtained based on the uptake time.
Sibille teaches at Paragraph 0057 that the attention mask can be obtained based on a SUV threshold in the fast SPECT image (associated with a second uptake time). For instance, a SUV threshold may be selected and the fast SPECT image or input image may be filtered based on the threshold. In some cases, the filtered fast SPECT image may be used as the attention mask. The attention mask may need more accurate boundary enhancement compared to the normal structures and background. Including the attention mask may beneficially preserve the quantification accuracy and/or sharpness of the features (e.g., features of lesion, segment of bone lesion). The attention mask can be obtained using any other suitable methods such as utilizing deep learning techniques. Alternatively, the attention mask may be generated manually by manual selection/segmentation on an input image.
Sibille teaches at Paragraph 0085 that the training datasets for may comprise pairs of standard acquisition (standard uptake time) and shortened acquisition SPECT images (shortened uptake time), CT images and/or attention feature map from the same subject.
Chen teaches that the features associated with the SUV values as well as the uptake time of the predicted PET parameter image are modified.
Chen teaches at Paragraph 0041 that when the original PET parameter image is the K1 parameter image, a maximum acquisition time corresponding to the preset acquisition time range is the total acquisition duration corresponding to the dynamic PET image set and at Paragraph 0055 the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained and the input image is input into the image enhancement model to obtain the output predicted PET parameter image…the image quality of the PET parameter image is improved while preserving image details of the PET parameter image.
Chen teaches at Paragraph 0057 that the original PET parameter image is determined based on an obtained dynamic PET image set and at Paragraph 0086 that the input image is a dynamic PET image corresponding a preset acquisition time range (uptake time of 5 minutes) in the dynamic PET image set and at Paragraph 0087 that the predicted PET parameter image is used as the target PET parameter image corresponding to the original PET parameter image so that the problem that an existing neural network model method requires preparation of a high-quality PET parameter image is solved.
Chen teaches at Paragraph 0088, when the original PET parameter image is a K1 parameter image, the maximum acquisition time corresponding to the preset acquisition time range is (60 minutes) a total acquisition time duration corresponding to the dynamic PET image set and at Paragraph 0099 determining a Euclidean distance difference between the original PET parameter image and the predicted PET parameter image based on the L2 loss function (the predicted PET parameter image has the same up take time as the original PET parameter image which is 60 minutes).
Chen teaches at Paragraph 0081 that at least two parameter feature maps are output based on the inputted input image by at least two encoding convolutional networks in the encoder and the predicted PET parameter image is outputted based on the at least two parameter feature maps
Chen teaches at Paragraph 0047 that the input image is input into an image enhancement model to obtain an output predicted PET parameter image and at Paragraph 0048 that a model architecture of the image enhancement model includes a generative adversarial network architecture, a U-Net architecture, super resolution convolutional neural networks. Chen teaches at Paragraph 0049 that a model parameter of the image enhancement model is adjusted based on the original PET parameter image and predicted PET parameter image.
Chen teaches at Paragraph 0055 that the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained, wherein the input image is the noise image, the dynamic PET image corresponding the preset acquisition time range in the dynamic PET image set or the dynamic SUV image corresponding to the dynamic PET image, the input image is input into the image enhancement model to obtain the output predicted PET parameter image, the model parameter of the image enhancement model is adjusted based on the original PET parameter image and the predicted PET parameter image until the preset number of iterations is met.
Xiang teaches at Paragraph 0013 that the deep learning network model is trained using training data comprising a SPECT image acquired using shortened acquisition time, a corresponding CT image and a SPECT image acquired using a standard acquisition time.
Xiang teaches at Paragraph 0009 that the term “fast scan SPECT image” may generally refer to a SPECT image acquired under shortened acquisition time at an acceleration factor of a value greater than 1. The standard acquisition time is at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20 times of the shortened acquisition time of the fast scan SPECT image.
Xiang teaches at Paragraph 0031 that the present disclosure provides an imaging acceleration method employing deep learning techniques to improve the image quality acquired with shortened acquisition time, i.e., fast scan and the deep learning techniques may comprise using multi-modality and multi-scale feature aggregation-based framework to generate SPECT images from a fast scan with image quality comparable to SPECT images acquired with standard acquisition time. The methods and systems herein may effectively reduce the radiation exposure by shortening the SPECT/CT examination time, e.g., shorten the SPECT scan time).
Siddu teaches at Paragraph 0090 that PET normalization helps to normalize SUV values that will vary with respect to the uptake time of the nuclear tracer and multi-modal fusion of hand-crafted features from CT and PET images such as radiomics features extracted from an ROI bounding box is proposed as well as a machine learning model to map Gleason scores.
Chen teaches at Paragraph 0071 normalizing the input image to obtain a normalized input image and registering the normalized input image with the original PET image to obtain a registered input image. Sibille teaches at Paragraph 0010 that the quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contract and the accuracy, e.g., standardized uptake value (SUV) in lesion regions and the method herein may boost the image contrast and the standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention maps.
According to Siddu, the normalized input image of Chen includes the normalized SUV values as a function of uptake time where Sibille teaches SUV values (features of the enhanced SPECT image) are boosted up to the standard uptake time.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have boosted features of an enhanced PET image as a function of uptake time up to the quality of the PET image acquired at the standard uptake time. One of the ordinary skill in the art would have encoded the features as function of uptake time in a series of encoding convolutional networks.
Re Claim 3:
The claim 3 encompasses the same scope of invention as that of the claim 2 except additional claim limitation that inputting the measurement data to a neural network, the neural network configured to generate the feature maps based on the measurement data.
Chen and Sibille/Xiang further teach the claim limitation that inputting the measurement data to a neural network, the neural network configured to generate the feature maps based on the measurement data (
Sibille teaches at Paragraph 0037 that new SUV value is modified from the standard SUV value based on the uptake time data to provide more accurate SUV measures based on the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image using standard acquisition time as reference where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Chen teaches at Paragraph 0041 that when the original PET parameter image is the K1 parameter image, a maximum acquisition time corresponding to the preset acquisition time range is the total acquisition duration corresponding to the dynamic PET image set and at Paragraph 0055 the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained and the input image is input into the image enhancement model to obtain the output predicted PET parameter image…the image quality of the PET parameter image is improved while preserving image details of the PET parameter image.
Chen teaches at Paragraph 0057 that the original PET parameter image is determined based on an obtained dynamic PET image set and at Paragraph 0086 that the input image is a dynamic PET image corresponding a preset acquisition time range (uptake time of 5 minutes) in the dynamic PET image set and at Paragraph 0087 that the predicted PET parameter image is used as the target PET parameter image corresponding to the original PET parameter image so that the problem that an existing neural network model method requires preparation of a high-quality PET parameter image is solved.
Chen teaches at Paragraph 0088, when the original PET parameter image is a K1 parameter image, the maximum acquisition time corresponding to the preset acquisition time range is (60 minutes) a total acquisition time duration corresponding to the dynamic PET image set and at Paragraph 0099 determining a Euclidean distance difference between the original PET parameter image and the predicted PET parameter image based on the L2 loss function (the predicted PET parameter image has the same up take time as the original PET parameter image which is 60 minutes).
Chen teaches at Paragraph 0081 that at least two parameter feature maps are output based on the inputted input image by at least two encoding convolutional networks in the encoder and the predicted PET parameter image is outputted based on the at least two parameter feature maps.
Chen teaches at Paragraph 0047 that the input image is input into an image enhancement model to obtain an output predicted PET parameter image and at Paragraph 0048 that a model architecture of the image enhancement model includes a generative adversarial network architecture, a U-Net architecture, super resolution convolutional neural networks. Chen teaches at Paragraph 0049 that a model parameter of the image enhancement model is adjusted based on the original PET parameter image and predicted PET parameter image.
Chen teaches at Paragraph 0055 that the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained, wherein the input image is the noise image, the dynamic PET image corresponding the preset acquisition time range in the dynamic PET image set or the dynamic SUV image corresponding to the dynamic PET image, the input image is input into the image enhancement model to obtain the output predicted PET parameter image, the model parameter of the image enhancement model is adjusted based on the original PET parameter image and the predicted PET parameter image until the preset number of iterations is met.
Xiang teaches at Paragraph 0013 that the deep learning network model is trained using training data comprising a SPECT image acquired using shortened acquisition time, a corresponding CT image and a SPECT image acquired using a standard acquisition time.
Xiang teaches at Paragraph 0009 that the term “fast scan SPECT image” may generally refer to a SPECT image acquired under shortened acquisition time at an acceleration factor of a value greater than 1. The standard acquisition time is at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20 times of the shortened acquisition time of the fast scan SPECT image.
Xiang teaches at Paragraph 0031 that the present disclosure provides an imaging acceleration method employing deep learning techniques to improve the image quality acquired with shortened acquisition time, i.e., fast scan and the deep learning techniques may comprise using multi-modality and multi-scale feature aggregation-based framework to generate SPECT images from a fast scan with image quality comparable to SPECT images acquired with standard acquisition time. The methods and systems herein may effectively reduce the radiation exposure by shortening the SPECT/CT examination time, e.g., shorten the SPECT scan time).
Re Claim 4:
The claim 4 encompasses the same scope of invention as that of the claim 3 except additional claim limitation that the neural network comprises an encoder and a modulator, the computer-implemented method further comprising:
inputting the measurement data to the encoder;
generating, by the encoder, the feature maps; and
modifying, by the modulator, the feature maps based on the uptake time data.
Chen and Sibille/Xiang further teach the claim limitation that the neural network comprises an encoder and a modulator, the computer-implemented method further comprising:
inputting the measurement data to the encoder;
generating, by the encoder, the feature maps; and
modifying, by the modulator, the feature maps based on the uptake time data (
Sibille teaches at Paragraph 0037 that new SUV value is modified from the standard SUV value as a function of the enhanced SPECT image associated with the second uptake time data to provide more accurate SUV measures by minimizing the difference in the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image associated with the standard acquisition time where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Chen teaches at Paragraph 0060 that the encoder contains at least two encoding convolutional networks and at Paragraph 0061 that at least two parameter feature maps are output based on the inputted input image by the at least two encoding convolutional networks in the encoder.
Chen teaches at Paragraph [0064] the first parameter feature map is determined based on the inputted input image (associated with a first uptake time) by the first encoding convolutional network (i=1) in the encoder,
Chen teaches at Paragraph 0068 that, by the first decoding convolutional network (j=1) in the decoder, a first up-sampling feature map is determined based on the last parameter feature map output by the last encoding convolutional network in the encoder, and the first up-sampling feature map is output to a first bilinear interpolation layer. A first interpolation feature vector is determined based on the first upsampling feature map by the first bilinear interpolation layer in the decoder, and the first interpolation feature map is output to the second decoding convolutional network. By a current decoding convolutional network (1<j<n) in the decoder, an i.sup.th upsampling feature map is determined based on an (i−1).sup.th interpolation feature map output by an (i−1).sup.th bilinear interpolation layer as well as the parameter feature map input by the encoding convolutional network (i=n−j+1) in the encoder that corresponds to the current decoding convolutional network, and the i.sup.th upsampling feature map is output to an i.sup.th bilinear interpolation layer. By analogy, by a last decoding convolutional network (j=n) in the decoder, the predicted PET parameter image is determined based on an (n−1).sup.th interpolation feature map output by an (n−1).sup.th bilinear interpolation layer as well as the first parameter feature map input (associated with the first uptake time) by the first encoding convolutional network in the encoder, and the predicted PET parameter image is output (associated with the second uptake time).
Chen teaches at Paragraph 0041 that when the original PET parameter image is the K1 parameter image, a maximum acquisition time corresponding to the preset acquisition time range is the total acquisition duration corresponding to the dynamic PET image set and at Paragraph 0055 the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained and the input image is input into the image enhancement model to obtain the output predicted PET parameter image…the image quality of the PET parameter image is improved while preserving image details of the PET parameter image.
Chen teaches at Paragraph 0057 that the original PET parameter image is determined based on an obtained dynamic PET image set and at Paragraph 0086 that the input image is a dynamic PET image corresponding a preset acquisition time range (uptake time of 5 minutes) in the dynamic PET image set and at Paragraph 0087 that the predicted PET parameter image is used as the target PET parameter image corresponding to the original PET parameter image so that the problem that an existing neural network model method requires preparation of a high-quality PET parameter image is solved.
Chen teaches at Paragraph 0088, when the original PET parameter image is a K1 parameter image, the maximum acquisition time corresponding to the preset acquisition time range is (60 minutes) a total acquisition time duration corresponding to the dynamic PET image set and at Paragraph 0099 determining a Euclidean distance difference between the original PET parameter image and the predicted PET parameter image based on the L2 loss function (the predicted PET parameter image has the same up take time as the original PET parameter image which is 60 minutes).
Chen teaches at Paragraph 0081 that at least two parameter feature maps are output based on the inputted input image by at least two encoding convolutional networks in the encoder and the predicted PET parameter image is outputted based on the at least two parameter feature maps.
Chen teaches at Paragraph 0047 that the input image is input into an image enhancement model to obtain an output predicted PET parameter image and at Paragraph 0048 that a model architecture of the image enhancement model includes a generative adversarial network architecture, a U-Net architecture, super resolution convolutional neural networks. Chen teaches at Paragraph 0049 that a model parameter of the image enhancement model is adjusted based on the original PET parameter image and predicted PET parameter image.
Chen teaches at Paragraph 0055 that the input image corresponding to the original PET parameter image determined based on the dynamic PET image set is obtained, wherein the input image is the noise image, the dynamic PET image corresponding the preset acquisition time range in the dynamic PET image set or the dynamic SUV image corresponding to the dynamic PET image, the input image is input into the image enhancement model to obtain the output predicted PET parameter image, the model parameter of the image enhancement model is adjusted based on the original PET parameter image and the predicted PET parameter image until the preset number of iterations is met.
Xiang teaches at Paragraph 0013 that the deep learning network model is trained using training data comprising a SPECT image acquired using shortened acquisition time, a corresponding CT image and a SPECT image acquired using a standard acquisition time.
Xiang teaches at Paragraph 0009 that the term “fast scan SPECT image” may generally refer to a SPECT image acquired under shortened acquisition time at an acceleration factor of a value greater than 1. The standard acquisition time is at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20 times of the shortened acquisition time of the fast scan SPECT image.
Xiang teaches at Paragraph 0031 that the present disclosure provides an imaging acceleration method employing deep learning techniques to improve the image quality acquired with shortened acquisition time, i.e., fast scan and the deep learning techniques may comprise using multi-modality and multi-scale feature aggregation-based framework to generate SPECT images from a fast scan with image quality comparable to SPECT images acquired with standard acquisition time. The methods and systems herein may effectively reduce the radiation exposure by shortening the SPECT/CT examination time, e.g., shorten the SPECT scan time).
Siddu teaches at Paragraph 0090 that PET normalization helps to normalize SUV values that will vary with respect to the uptake time of the nuclear tracer and multi-modal fusion of hand-crafted features from CT and PET images such as radiomics features extracted from an ROI bounding box is proposed as well as a machine learning model to map Gleason scores.
Chen teaches at Paragraph 0071 normalizing the input image to obtain a normalized input image and registering the normalized input image with the original PET image to obtain a registered input image. Sibille teaches at Paragraph 0010 that the quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contrast and the accuracy, e.g., standardized uptake value (SUV) in lesion regions and the method herein may boost the image contrast and the standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention maps.
It is noted that the SUV value is directly associated with the uptake time. Thus the new SPECT image has the adjusted uptake time when the image contrast and accuracy is improved and the uptake time is improved up to the standard uptake time associated with the standardized uptake value (SUV).
According to Siddu, the normalized input image of Chen includes the normalized SUV values as a function of uptake time where Sibille teaches SUV values (features of the enhanced SPECT image) are boosted up to the standard uptake time.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have boosted features of an enhanced PET image as a function of uptake time up to the quality of the PET image acquired at the standard uptake time. One of the ordinary skill in the art would have encoded the features as function of uptake time in a series of encoding convolutional networks.
Re Claim 5:
The claim 5 encompasses the same scope of invention as that of the claim 4 except additional claim limitation that the neural network comprises a decoder, the computer-implemented method further comprising: receiving, by the decoder, the modified feature maps; and generating, by the decoder, the output image data based on the modified feature maps.
Chen further teaches the claim limitation that the neural network comprises a decoder, the computer-implemented method further comprising: receiving, by the decoder, the modified feature maps; and generating, by the decoder, the output image data based on the modified feature maps (Chen teaches at Paragraph 0060 that the encoder contains at least two encoding convolutional networks and at Paragraph 0061 that at least two parameter feature maps are output based on the inputted input image by the at least two encoding convolutional networks in the encoder.
Chen teaches at Paragraph 0068 that, by the first decoding convolutional network (j=1) in the decoder, a first up-sampling feature map is determined based on the last parameter feature map output by the last encoding convolutional network in the encoder, and the first up-sampling feature map is output to a first bilinear interpolation layer. A first interpolation feature vector is determined based on the first upsampling feature map by the first bilinear interpolation layer in the decoder, and the first interpolation feature map is output to the second decoding convolutional network. By a current decoding convolutional network (1<j<n) in the decoder, an i.sup.th upsampling feature map is determined based on an (i−1).sup.th interpolation feature map output by an (i−1).sup.th bilinear interpolation layer as well as the parameter feature map input by the encoding convolutional network (i=n−j+1) in the encoder that corresponds to the current decoding convolutional network, and the i.sup.th upsampling feature map is output to an i.sup.th bilinear interpolation layer. By analogy, by a last decoding convolutional network (j=n) in the decoder, the predicted PET parameter image is determined based on an (n−1).sup.th interpolation feature map output by an (n−1).sup.th bilinear interpolation layer as well as the first parameter feature map input by the first encoding convolutional network in the encoder, and the predicted PET parameter image is output).
Re Claim 6:
The claim 6 encompasses the same scope of invention as that of the claim 5 except additional claim limitation that receiving, by the decoder, encoded features from the encoder; and generating, by the decoder, the output image data based on the encoded features.
Chen further teaches the claim limitation that receiving, by the decoder, encoded features from the encoder; and generating, by the decoder, the output image data based on the encoded features (Chen teaches at Paragraph 0060 that the encoder contains at least two encoding convolutional networks and at Paragraph 0061 that at least two parameter feature maps are output based on the inputted input image by the at least two encoding convolutional networks in the encoder.
Chen teaches at Paragraph 0068 that, by the first decoding convolutional network (j=1) in the decoder, a first up-sampling feature map is determined based on the last parameter feature map output by the last encoding convolutional network in the encoder, and the first up-sampling feature map is output to a first bilinear interpolation layer. A first interpolation feature vector is determined based on the first upsampling feature map by the first bilinear interpolation layer in the decoder, and the first interpolation feature map is output to the second decoding convolutional network. By a current decoding convolutional network (1<j<n) in the decoder, an i.sup.th upsampling feature map is determined based on an (i−1).sup.th interpolation feature map output by an (i−1).sup.th bilinear interpolation layer as well as the parameter feature map input by the encoding convolutional network (i=n−j+1) in the encoder that corresponds to the current decoding convolutional network, and the i.sup.th upsampling feature map is output to an i.sup.th bilinear interpolation layer. By analogy, by a last decoding convolutional network (j=n) in the decoder, the predicted PET parameter image is determined based on an (n−1).sup.th interpolation feature map output by an (n−1).sup.th bilinear interpolation layer as well as the first parameter feature map input by the first encoding convolutional network in the encoder, and the predicted PET parameter image is output).
Re Claim 7:
The claim 7 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that the second uptake time is greater than the first uptake time.
Sibille further teaches the claim limitation that the second uptake time is greater than the first uptake time (
Sibille teaches at Paragraph 0037 that new SUV value is modified from the standard SUV value based on the uptake time data to provide more accurate SUV measures based on the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image using standard acquisition time as reference where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Re Claim 8:
The claim 8 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that applying the trained machine learning process to the measurement data and the uptake time data further comprises: generating uptake time vectors based on the uptake time data; inputting the uptake time vectors to a neural network; and based on inputting the uptake time vectors to the neural network, generating uptake time embeddings.
Chen in view of Sibille further teaches the claim limitation that applying the trained machine learning process to the measurement data and the uptake time data further comprises: generating uptake time vectors based on the uptake time data; inputting the uptake time vectors to a neural network; and based on inputting the uptake time vectors to the neural network, generating uptake time embeddings (Sibille teaches at Paragraph 0037 that new SUV value is modified from the standard SUV value as a function of the enhanced SPECT image associated with the second uptake time data to provide more accurate SUV measures by minimizing the difference in the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image associated with the standard acquisition time where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Chen teaches at Paragraph 0060 that the encoder contains at least two encoding convolutional networks and at Paragraph 0061 that at least two parameter feature maps are output based on the inputted input image by the at least two encoding convolutional networks in the encoder.
Chen teaches at Paragraph 0064 that the first parameter feature map is determined based on the input image (associated with the input uptake time) by the first encoding convolutional network (to generate feature map embedding).
Chen teaches at Paragraph 0068 that, by the first decoding convolutional network (j=1) in the decoder, a first up-sampling feature map is determined based on the last parameter feature map output by the last encoding convolutional network in the encoder, and the first up-sampling feature map is output to a first bilinear interpolation layer. A first interpolation feature vector is determined based on the first upsampling feature map by the first bilinear interpolation layer in the decoder, and the first interpolation feature map is output to the second decoding convolutional network. By a current decoding convolutional network (1<j<n) in the decoder, an i.sup.th upsampling feature map is determined based on an (i−1).sup.th interpolation feature map output by an (i−1).sup.th bilinear interpolation layer as well as the parameter feature map input by the encoding convolutional network (i=n−j+1) in the encoder that corresponds to the current decoding convolutional network, and the i.sup.th upsampling feature map is output to an i.sup.th bilinear interpolation layer. By analogy, by a last decoding convolutional network (j=n) in the decoder, the predicted PET parameter image is determined based on an (n−1).sup.th interpolation feature map output (associated with the output uptake time) by an (n−1).sup.th bilinear interpolation layer as well as the first parameter feature map input by the first encoding convolutional network in the encoder, and the predicted PET parameter image is output).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have included uptake time or SUV value into the feature map at each stage of the encoding convolutional network of the n encoding convolutional networks to have characterized the ROI based on the n number of uptake times or SUV values of the feature maps of the n encoding convolutional networks. One of the ordinary skill in the art would have been motivated to have generated enhanced medical image as a function of the uptake times or SUV values of the n encoding convolutional networks.
Siddu teaches at Paragraph 0090 that PET normalization helps to normalize SUV values that will vary with respect to the uptake time of the nuclear tracer and multi-modal fusion of hand-crafted features from CT and PET images such as radiomics features extracted from an ROI bounding box is proposed as well as a machine learning model to map Gleason scores.
Chen teaches at Paragraph 0071 normalizing the input image to obtain a normalized input image and registering the normalized input image with the original PET image to obtain a registered input image. Sibille teaches at Paragraph 0010 that the quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contract and the accuracy, e.g., standardized uptake value (SUV) in lesion regions and the method herein may boost the image contrast and the standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention maps.
According to Siddu, the normalized input image of Chen includes the normalized SUV values as a function of uptake time where Sibille teaches SUV values (features of the enhanced SPECT image) are boosted up to the standard uptake time.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have boosted features of an enhanced PET image as a function of uptake time up to the quality of the PET image acquired at the standard uptake time. One of the ordinary skill in the art would have encoded the features as function of uptake time in a series of encoding convolutional networks.
Re Claim 9:
The claim 9 encompasses the same scope of invention as that of the claim 8 except additional claim limitation that generating maps based on the measurement data; and
generating modified feature maps based on the uptake time embeddings.
Chen in view of Sibille further teaches the claim limitation that generating maps based on the measurement data; and
generating modified feature maps based on the uptake time embeddings (
Sibille teaches at Paragraph 0037 that new SUV value is modified from the standard SUV value as a function of the enhanced SPECT image associated with the second uptake time data to provide more accurate SUV measures by minimizing the difference in the attention feature maps where the enhanced medical image is obtained by calculating the attention loss and comparing the attention feature map of the enhanced medical image with the attention feature map of the SPECT image associated with the standard acquisition time where the attention feature map of the enhanced medical image (associated with high standard uptake time) is modified from the attention feature map of the fast-scanned SPECT image (associated with low uptake time).
Chen teaches at Paragraph 0060 that the encoder contains at least two encoding convolutional networks and at Paragraph 0061 that at least two parameter feature maps are output based on the inputted input image by the at least two encoding convolutional networks in the encoder.
Chen teaches at Paragraph 0064 that the first parameter feature map is determined based on the input image (associated with the input uptake time) by the first encoding convolutional network (to generate feature map embedding).
Chen teaches at Paragraph 0068 that, by the first decoding convolutional network (j=1) in the decoder, a first up-sampling feature map is determined based on the last parameter feature map output by the last encoding convolutional network in the encoder, and the first up-sampling feature map is output to a first bilinear interpolation layer. A first interpolation feature vector is determined based on the first upsampling feature map by the first bilinear interpolation layer in the decoder, and the first interpolation feature map is output to the second decoding convolutional network. By a current decoding convolutional network (1<j<n) in the decoder, an i.sup.th upsampling feature map is determined based on an (i−1).sup.th interpolation feature map output by an (i−1).sup.th bilinear interpolation layer as well as the parameter feature map input by the encoding convolutional network (i=n−j+1) in the encoder that corresponds to the current decoding convolutional network, and the i.sup.th upsampling feature map is output to an i.sup.th bilinear interpolation layer. By analogy, by a last decoding convolutional network (j=n) in the decoder, the predicted PET parameter image is determined based on an (n−1).sup.th interpolation feature map output (associated with the output uptake time) by an (n−1).sup.th bilinear interpolation layer as well as the first parameter feature map input by the first encoding convolutional network in the encoder, and the predicted PET parameter image is output).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have included uptake time or SUV value into the feature map of Chen to have characterized the ROI based on the uptake time or SUV value of the feature map. One of the ordinary skill in the art would have been motivated to have generated enhanced medical image as a function of the uptake time or SUV value.
Siddu teaches at Paragraph 0090 that PET normalization helps to normalize SUV values that will vary with respect to the uptake time of the nuclear tracer and multi-modal fusion of hand-crafted features from CT and PET images such as radiomics features extracted from an ROI bounding box is proposed as well as a machine learning model to map Gleason scores.
Chen teaches at Paragraph 0071 normalizing the input image to obtain a normalized input image and registering the normalized input image with the original PET image to obtain a registered input image. Sibille teaches at Paragraph 0010 that the quantification accuracy may be preserved or improved over the fast scan SPECT image by boosting the image contract and the accuracy, e.g., standardized uptake value (SUV) in lesion regions and the method herein may boost the image contrast and the standardized uptake value (SUV) in lesion regions by penalizing the losses in these regions such as utilizing corresponding attention maps.
According to Siddu, the normalized input image of Chen includes the normalized SUV values as a function of uptake time where Sibille teaches SUV values (features of the enhanced SPECT image) are boosted up to the standard uptake time.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have boosted features of an enhanced PET image as a function of uptake time up to the quality of the PET image acquired at the standard uptake time. One of the ordinary skill in the art would have encoded the features as function of uptake time in a series of encoding convolutional networks.
Re Claim 10:
The claim 10 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that providing the output image for display.
Sibille teaches the claim limitation that providing the output image for display (Sibille teaches at Paragraph 0087 that the UI may display the improved SPECT image).
Re Claim 11:
The claim 11 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that the trained machine learning process is based on a trained convolutional neural network.
Sibille further teaches the claim limitation that the trained machine learning process is based on a trained convolutional neural network (Sibille teaches at Paragraph [0007] The present disclosure provides an acceleration method employing deep learning techniques to improve the image quality acquired with shortened acquisition time (i.e., fast scan). In some cases, the deep learning techniques may comprise using a convolutional neural network (CNN)-based framework to generate SPECT images from a fast scan with image quality comparable to SPECT images acquired with standard acquisition time).
The Claim 12 recites a non-transitory computer readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations comprising:
receiving measurement data characterizing a scanned image of a subject;
receiving uptake time data characterizing a first uptake time of the scanned image;
applying a trained machine learning process to the measurement data and the uptake time data and, based on the application of the trained machine learning process to the measurement data and the uptake time data, generating output image data characterizing an output image at a second uptake time; and
storing the output image in a data repository.
The claim 12 is in parallel with the claim 1 in the form of a computer program product. The claim 12 is subject to the same rationale of rejection as the claim 1.
Moreover, Chen further teaches a non-transitory computer readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform operations [of the claim 1] (Chen teaches at Paragraph 0020 a computer readable storage medium stores computer instruction, when executed, used to cause a processor to implement the method for enhancing the PET parameter image).
Re Claim 13:
The claim 13 encompasses the same scope of invention as that of the claim 12 except additional claim limitation that storing further instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising:
generating feature maps based on the measurement data; and
modifying the feature maps based on the uptake time data.
The claim 13 is in parallel with the claim 2 in the form of a computer program product. The claim 13 is subject to the same rationale of rejection as the claim 2.
Re Claim 14:
The claim 14 encompasses the same scope of invention as that of the claim 13 except additional claim limitation that storing further instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising inputting the measurement data to a neural network, the neural network configured to generate the feature maps based on the measurement data.
The claim 14 is in parallel with the claim 3 in the form of a computer program product. The claim 14 is subject to the same rationale of rejection as the claim 3.
Re Claim 15:
The claim 15 encompasses the same scope of invention as that of the claim 14 except additional claim limitation that the neural network comprises an encoder and a modulator, and wherein the non-transitory computer readable medium is storing further instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising: inputting the measurement data to the encoder; generating, by the encoder, the feature maps; and modifying, by the modulator, the feature maps based on the uptake time data.
The claim 15 is in parallel with the claim 4 in the form of a computer program product. The claim 15 is subject to the same rationale of rejection as the claim 4.
Re Claim 16:
The claim 16 encompasses the same scope of invention as that of the claim 15 except additional claim limitation that the neural network comprises a decoder, and wherein the non-transitory computer readable medium is storing instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising:
receiving, by the decoder, the modified feature maps; and
generating, by the decoder, the output image data based on the modified feature maps.
The claim 16 is in parallel with the claim 5 in the form of a computer program product. The claim 16 is subject to the same rationale of rejection as the claim 5.
Re Claim 17:
The claim 17 encompasses the same scope of invention as that of the claim 16 except additional claim limitation that storing further instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising:
receiving, by the decoder, encoded features from the encoder; and
generating, by the decoder, the output image data based on the encoded features.
The claim 17 is in parallel with the claim 6 in the form of a computer program product. The claim 17 is subject to the same rationale of rejection as the claim 6.
Re Claim 18:
The claim 18 encompasses the same scope of invention as that of the claim 12 except additional claim limitation that storing further instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising:
generating uptake time vectors based on the uptake time data;
inputting the uptake time vectors to a neural network; and
based on inputting the uptake time vectors to the neural network, generating uptake time embeddings.
The claim 18 is in parallel with the claim 8 in the form of a computer program product. The claim 18 is subject to the same rationale of rejection as the claim 8.
Re Claim 19:
The claim 19 encompasses the same scope of invention as that of the claim 18 except additional claim limitation that storing further instructions that, when executed by the at least one processor, further cause the at least one processor to perform operations comprising:
generating maps based on the measurement data; and
generating modified feature maps based on the uptake time embeddings.
The claim 19 is in parallel with the claim 9 in the form of a computer program product. The claim 19 is subject to the same rationale of rejection as the claim 9.
Re Claim 20:
The claim 20 recites a system comprising:
a data repository; and
at least one processor communicatively coupled to the data repository, the at least one processor configured to:
receive measurement data characterizing a scanned image of a subject; receive uptake time data characterizing a first uptake time of the scanned image;
apply a trained machine learning process to the measurement data and the uptake time data and, based on the application of the trained machine learning process to the measurement data and the uptake time data, generating output image data characterizing an output image at a second uptake time; and
store the output image in the data repository.
The claim 20 is in parallel with the claim 1 in the apparatus form. The claim 20 is subject to the same rationale of rejection as the claim 1.
Moreover, Chen further teaches a system comprising:
a data repository; and
at least one processor communicatively coupled to the data repository, the at least one processor configured to [perform the operations of the claim 1] (Chen teaches at Paragraph 0020 a computer readable storage medium stores computer instruction, when executed, used to cause a processor to implement the method for enhancing the PET parameter image).
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JIN CHENG WANG whose telephone number is (571)272-7665. The examiner can normally be reached Mon-Fri 8:00-5:00.
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, King Poon can be reached at 571-270-0728. 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.
/JIN CHENG WANG/Primary Examiner, Art Unit 2617