# US 10,769,785 B2 · Neural network for neurodegenerative disease classification > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,769,785 B2 - Document ID: US10769785B2 - Status: Granted - Kind: B2 - Application number: 16/238,696 - Filed: 2019-01-03 - Published: 2020-03-19 - Granted: 2020-09-08 - Earliest claimed: 2018-09-14 - Assignees: Siemens Medical Solutions USA, Inc. - Applicants: Siemens Medical Solutions USA, Inc. - Inventors: Rachid Fahmi; Sven Zuehlsdorff; Ludovic Sibille - Web page: https://patentgrep.com/patent/US10769785B2 ## CPC - G06T 7/0012: {Biomedical image inspection} - A61B 6/037: {Emission tomography} - G06N 3/045: Combinations of networks - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G06T 7/11: Region-based segmentation - G06T 2207/10104: Positron emission tomography [PET] - G06T 2207/10108: Single photon emission computed tomography [SPECT] - G06T 2207/20081: Training; Learning - G06T 2207/20084: Artificial neural networks [ANN] - G06T 2207/20132: Image cropping - G06T 2207/30016: Brain - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A method for configuring a neural network comprises: accessing a plurality of three-dimensional (3D) emission image data sets collected by an emission scanner from respective brains of respective subjects; transforming each of the plurality of 3D emission image data sets to a respective two-dimensional (2D) image; cropping portions of each respective 2D image to remove image data corresponding to tissue outside of a striatum of each of the respective brains, to form respective cropped 2D striatum images; and training a neural network to detect a presence of a Parkinsonian syndrome using the cropped 2D striatum images. ## Claims ### Claim 1 A method for configuring a neural network, comprising: accessing a plurality of three-dimensional (3D) emission image data sets collected by an emission scanner from respective brains of respective subjects; transforming each of the plurality of 3D emission image data sets to a respective two-dimensional (2D) image; cropping portions of each respective 2D image to remove image data corresponding to tissue outside of a striatum of each of the respective brains, to form respective cropped 2D striatum images; and training a neural network to detect a presence of a Parkinsonian syndrome using the cropped 2D striatum images. ### Claim 2 (depends on 1) The method of claim 1, wherein the scanner is a positron emission tomography (PET) or a single-photon emission computerized tomography (SPECT) scanner. ### Claim 3 (depends on 1) The method of claim 1, further comprising spatially normalizing the 3D emission image data sets to a standardized space before the transforming. ### Claim 4 (depends on 3) The method of claim 3, further comprising normalizing intensity of the 3D emission image data sets before the transforming. ### Claim 5 (depends on 1) The method of claim 1, wherein the 3D emission image data sets correspond to subjects injected with a dopaminergic radioligand. ### Claim 6 (depends on 1) The method of claim 1, wherein the Parkinsonian syndrome is Parkinson's disease, multiple system atrophy and progressive supranuclear palsy. ### Claim 7 (depends on 6) The method of claim 6, further comprising: injecting each of the subjects with a tracer targeting dopaminergic system of the subjects; and scanning the subjects to collect the 3D emission image data prior to the accessing step. ### Claim 8 (depends on 7) The method of claim 7, wherein the tracer comprises one of the group consisting of: [123I]FP-CIT, [123I]β-CIT, [99mTc]TRODAT-1, [123I]PE2I, [123I]-altropane, [123I]IBZM, [18F]dopa, [11C]dopa, [11C]cocaine [3H]WIN, [11C]altropane, [11C]/[18F]β-CFT, [11C]FE-CIT, [11C]dMP, [11C]DTBZ, [11C]raclopride, [18F]DMFP, or [11C]NMSP. ### Claim 9 (depends on 1) The method of claim 1, wherein each 3D image data set has a plurality of slices, each slice has a plurality of voxels, each 2D image has a plurality of pixels, and the transforming step includes averaging corresponding voxels from at least two of the slices to determine a value of a corresponding pixel of the respective 2D image. ### Claim 10 (depends on 9) The method of claim 9, further comprising determining which six consecutive slices among the plurality of slices of one of the 3D emission image data sets have highest peak distribution volume ratios, wherein the averaging is performed on the six consecutive slices. ### Claim 11 (depends on 10) The method of claim 10, wherein each of the plurality of slices includes a respective cross-section of the striatum of one of the brains. ### Claim 12 (depends on 1) The method of claim 1, wherein the neural network is a convolutional neural network. ### Claim 13 A system for configuring a neural network, comprising: a processor coupled to access a plurality of three-dimensional (3D) emission image data sets collected by an emission scanner from respective brains of respective subjects; the processor configured to transform each of the plurality of 3D emission image data sets to a respective two-dimensional (2D) image; the processor configured to crop portions of each respective 2D image to remove image data corresponding to tissue outside of a striatum of each of the respective brains, to form respective cropped 2D striatum images; and a neural network configured to receive the cropped 2D striatum images for training the neural network to detect a presence of a Parkinsonian syndrome using the cropped 2D striatum images. ### Claim 14 (depends on 13) The system of claim 13, further comprising a positron emission tomography (PET) or a single-photon emission computerized tomography (SPECT) scanner to which the processor is coupled, for providing the plurality of 3D emission image data sets to the processor. ### Claim 15 (depends on 13) The system of claim 13, wherein the neural network is a convolutional neural network. ### Claim 16 A non-transitory, machine readable storage medium encoded with program instructions, such that when a processor executes the program instructions, the program instructions configure the processor to perform a method for configuring a neural network, comprising: accessing a plurality of three-dimensional (3D) emission image data sets collected by an emission scanner from respective brains of respective subjects; transforming each of the plurality of 3D emission image data sets to a respective two-dimensional (2D) image; cropping portions of each respective 2D image to remove image data corresponding to tissue outside of a striatum of each of the respective brains, to form respective cropped 2D striatum images; and training a neural network to detect a presence of a Parkinsonian syndrome using the cropped 2D striatum images. ### Claim 17 (depends on 16) The non-transitory, machine readable storage medium of claim 16, wherein the program instructions further configure the process for: spatially normalizing the 3D emission image data sets to a standardized space; and normalizing intensity of the spatially normalized 3D emission image data sets. ### Claim 18 (depends on 16) The non-transitory, machine readable storage medium of claim 16, wherein each 3D image data set has a plurality of slices, each slice has a plurality of voxels, each 2D image has a plurality of pixels, and the transforming step includes averaging corresponding voxels from at least two of the slices to determine a value of a corresponding pixel of the respective 2D image. ### Claim 19 (depends on 18) The non-transitory, machine readable storage medium of claim 18, wherein the program instructions further configure the process for determining which six consecutive slices among the plurality of slices of one of the 3D emission image data sets has a have highest peak distribution volume ratios, wherein the averaging is performed on the six consecutive slices. ### Claim 20 (depends on 16) The non-transitory, machine readable storage medium of claim 16, wherein the neural network is a convolutional neural network.