# US 12,181,552 B2 · Diffusion magnetic resonance imaging using spherical neural networks > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,181,552 B2 - Document ID: US12181552B2 - Status: Granted - Kind: B2 - Application number: 17/294,699 - Filed: 2019-11-19 - Published: 2022-01-13 - Granted: 2024-12-31 - Earliest claimed: 2018-11-20 - Assignees: Koninklijke Philips N.V. - Applicants: Koninklijke Philips N.V. - Inventors: Evan Schwab; Arne Ewald - Web page: https://patentgrep.com/patent/US12181552B2 ## CPC - G01R 33/5608: {Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels (image data processing or generation, in general G06T)} - G01R 33/56341: {Diffusion imaging} - G06N 3/02: Neural networks - G06V 10/764: using classification, e.g. of video objects - G06V 10/82: using neural networks - G06V 2201/03: Recognition of patterns in medical or anatomical images - G01R: MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES (indicating correct tuning of resonant circuits H03J3/12) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract The invention provides for a medical imaging system (100, 300). The medical imaging system comprises a memory (110) for storing machine executable instructions (120). The memory further contains an implementation of a trained convolutional neural network (122, 122′, 122″, 122′″, 122″″). The trained convolutional neural network comprises more than one spherical convolutional neural network portions (502, 502′). The trained convolutional neural network is configured for receiving diffusion magnetic resonance imaging data (124). The diffusion magnetic resonance imaging data comprises a spherical diffusion portion (500, 500′). The more than one spherical convolutional neural network portions are configured for receiving the spherical diffusion portion. The trained convolutional neural network comprises an output layer (508) configured for generating a neural network output (126) in response to inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network. The medical imaging system further comprises a processor (104) for controlling the machine executable instructions. Execution of the machine executable instructions causes the processor to: receive (200) the diffusion magnetic resonance imaging data; and generate (202) the neural network output by inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network. ## Claims ### Claim 1 A medical imaging system comprising: a memory for storing machine executable instructions wherein the memory further contains an implementation of a trained convolutional neural network, wherein the trained convolutional neural network comprises a spatial convolutional neural network portion and more than one spherical convolutional neural network portions, wherein the trained convolutional neural network is configured for receiving diffusion magnetic resonance imaging data wherein the diffusion magnetic resonance imaging data comprises a spatial portion and a spherical diffusion portion, wherein the spatial convolutional neural network portion is configured for receiving the spatial portion and the more than one spherical convolutional neural network portions are configured for receiving the spherical diffusion portion, wherein the trained convolutional neural network comprises an output layer configured for generating a neural network output in response to inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network; a processor for controlling the machine executable instructions, wherein execution of the machine executable instructions causes the processor to: receive the diffusion magnetic resonance imaging data; generate the neural network output by inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network wherein the neural network output comprises at least one of the following: a heatmap projected onto the diffusion magnetic resonance imaging data in a spatial and/or spherical domain; a heatmap projected onto a space of fiber tracts reconstructed using a tractography algorithm; or a global class prediction for the diffusion magnetic resonance imaging data at a global level, wherein the more than one spherical convolutional neural network portions and the spatial convolutional neural network portion are connected in parallel to the output layer, and wherein the more than one spherical convolutional neural network portions and the spatial convolutional neural network portion are interconnected via weight sharing, providing visualization of spherical and spatial activations of the trained convolutional neural network to highlight segments of the fiber tracts responsible for disease classification that overcomes limitations of manual feature analysis. ### Claim 2 (depends on 1) The medical imaging system of claim 1, wherein the diffusion magnetic resonance imaging data comprises voxels, wherein the diffusion magnetic resonance imaging data comprises the spherical diffusion portion for each of the voxels. ### Claim 3 (depends on 1) The medical imaging system of claim 1, wherein the spatial convolutional neural network portion and each of the more than one spherical convolutional neural network portions has an output that is each connected to an output concatenation, and wherein the output concatenation comprises the output layer. ### Claim 4 (depends on 1) The medical imaging system of claim 1, wherein the more than one spherical convolutional neural network portions comprise multiple spherical convolutional neural network portions, wherein the convolutional neural network comprises weight sharing between the multiple spherical convolutional neural network portions. ### Claim 5 (depends on 4) The medical imaging system of claim 4, wherein each of the multiple spherical convolutional neural network portions are measurements acquired at a different gradient magnetic field value. ### Claim 6 (depends on 1) The medical imaging system of claim 1, wherein the spherical diffusion portion is a diffusion signal for one or more gradient magnetic field values projected onto a sphere. ### Claim 7 (depends on 6) The medical imaging system of claim 6, wherein the diffusion magnetic resonance imaging data comprises any one of the following: diffusion tensor imaging data; angular resolution diffusion imaging data; multi shell angular resolution diffusion imaging data; and Q-Ball imaging data. ### Claim 8 (depends on 1) The medical imaging system of claim 1, wherein the spherical diffusion portion comprises a diffusivity model calculated from a diffusion signal projected onto a sphere. ### Claim 9 (depends on 8) The medical instrument of claim 8, wherein the diffusivity model is any one of the following: an orientation distribution function; a fiber orientation distribution function; and an ensemble average propagator restricted to two or more concentric spheres. ### Claim 10 (depends on 1) The medical imaging system claim 1, wherein the medical imaging system further comprises a magnetic resonance imaging system, wherein the memory further contains pulse sequence commands for controlling the magnetic resonance imaging system to acquire magnetic resonance acquisition data according to a diffusion weighted magnetic resonance imaging protocol, wherein execution of the machine executable instructions causes the processor to: control the magnetic resonance imaging system with the pulse sequence commands to acquire the magnetic resonance acquisition data; and reconstruct the diffusion magnetic resonance imaging data from the magnetic resonance acquisition data. ### Claim 11 A non-transitory computer readable medium comprising machine executable instructions stored therein for execution by a processor controlling a medical imaging system, wherein execution of the machine executable instructions causes the processor to: receive diffusion magnetic resonance imaging data; generate a neural network output by inputting the diffusion magnetic resonance imaging data into a trained convolutional neural network, wherein the trained convolutional neural network comprises a spatial convolutional neural network portion and more than one convolutional neural network portions, wherein the trained convolutional neural network is configured for receiving the diffusion magnetic resonance imaging data, wherein the diffusion magnetic resonance imaging data comprises a spatial portion and a spherical diffusion portion, wherein the spatial convolutional neural network portion is configured for receiving the spatial portion and the more than one spherical convolutional neural network portions are configured for receiving the spherical diffusion portion, wherein the trained convolutional neural network comprises an output layer configured for generating the neural network output in response to inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network wherein the neural network output comprises at least one of the following: a heatmap projected onto the diffusion magnetic resonance imaging data in a spatial and/or spherical domain; a heatmap projected onto a space of fiber tracts reconstructed using a tractography algorithm; or a global class prediction for the diffusion magnetic resonance imaging data at a global level, and wherein the more than one spherical convolutional neural network portions and the spatial convolutional neural network portion are connected in parallel to the output layer, and wherein the more than one spherical convolutional neural network portions and the spatial convolutional neural network portion are interconnected via weight sharing, providing visualization of spherical and spatial activations of the trained convolutional neural network to highlight segments of the fiber tracts responsible for disease classification that overcomes limitations of manual feature analysis. ### Claim 12 (depends on 11) The computer program product of claim 11, wherein the diffusion magnetic resonance imaging data comprises voxels, wherein the diffusion magnetic resonance imaging data comprises the spherical diffusion portion for each of the voxels. ### Claim 13 (depends on 11) The non-transitory computer readable medium of claim 11, wherein the spatial convolutional neural network portion and each of the more than one spherical convolutional neural network portions has an output that is each connected to an output concatenation, and wherein the output concatenation comprises the output layer. ### Claim 14 A method of medical image processing, wherein the method comprises: receiving diffusion magnetic resonance imaging data; generating the neural network output by inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network wherein the trained convolutional neural network comprises a spatial convolutional neural network portion and more than one spherical convolutional neural network portions, wherein the trained convolutional neural network is configured for receiving the diffusion magnetic resonance imaging data, wherein the diffusion magnetic resonance imaging data comprises a spatial portion and a spherical diffusion portion, wherein the spatial convolutional neural network portion is configured for receiving the spatial portion and the more than one spherical convolutional neural network portions are configured for receiving the spherical diffusion portion, wherein the trained convolutional neural network comprises an output layer configured for generating the neural network output in response to inputting the diffusion magnetic resonance imaging data into the trained convolutional neural network wherein the neural network output comprises at least one of the following: a heatmap projected onto the diffusion magnetic resonance imaging data in a spatial and/or spherical domain; a heatmap projected onto a space of fiber tracts reconstructed using a tractography algorithm; or a global class prediction for the diffusion magnetic resonance imaging data at a global level, and wherein the more than one spherical convolutional neural network portions and the spatial convolutional neural network portion are connected in parallel to the output layer, and wherein the more than one spherical convolutional neural network portions and the spatial convolutional neural network portion are interconnected via weight sharing, providing visualization of spherical and spatial activations of the trained convolutional neural network to highlight segments of the fiber tracts responsible for disease classification that overcomes limitations of manual feature analysis. ### Claim 15 (depends on 14) The method of medical image processing of claim 14, wherein the diffusion magnetic resonance imaging data comprises voxels, wherein the diffusion magnetic resonance imaging data comprises the spherical diffusion portion for each of the voxels. ### Claim 16 (depends on 14) The method of medical image processing of claim 14, wherein the spatial convolutional neural network portion and each of the more than one spherical convolutional neural network portions has an output that is each connected to an output concatenation, and wherein the output concatenation comprises the output layer.