Adaptive Template for Medical Image Spatial Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for spatial normalization of medical images, such as PET and SPECT, are inaccurate and costly due to reliance on expensive MRI or CT scans, and fail to account for individual variations in patient features.
Innovation Solution
A device using deep learning to generate an adaptive template for spatial normalization of medical images through a generative adversarial network, allowing for accurate normalization without the need for additional MRI or CT scans, by iteratively learning and minimizing differences with prestored MRI-based data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If MRI or CT images are captured together with functional images for spatial normalization, then spatial normalization accuracy is improved, but cost and time consumption increase
Solution Approach 1:
The patent creates a virtual copy (synthetic MRI image) from the functional image data itself using deep learning, eliminating the need for physical MRI/CT scanning. The GAN generates a synthetic structural image that mimics the appearance and anatomical features of a real MRI, which is then used for spatial normalization without requiring actual additional imaging procedures.
Solution Approach 2:
The functional image serves dual purposes: it is both the input data for analysis and the source material for generating the synthetic MRI image needed for normalization. The deep learning model extracts structural information directly from the functional image data, making the system self-sufficient without external imaging equipment.
2Measurement precision
If MRI or CT images are captured together with functional images for spatial normalization, then spatial normalization accuracy is improved, but time consumption increases
Solution Approach 1:
The patent creates a virtual copy (synthetic MRI image) from the functional image data itself using deep learning, eliminating the need for physical MRI/CT scanning. The GAN generates a synthetic structural image that mimics the appearance and anatomical features of a real MRI, which is then used for spatial normalization without requiring actual additional imaging procedures.
Solution Approach 2:
The functional image serves dual purposes: it is both the input data for analysis and the source material for generating the synthetic MRI image needed for normalization. The deep learning model extracts structural information directly from the functional image data, making the system self-sufficient without external imaging equipment.
3Productivity
If an average template obtained from various samples is used for spatial normalization, then processing speed is improved, but accuracy in reflecting individual patient features deteriorates
Solution Approach 1:
The patent transitions from a uniform average template to an adaptive template that preserves local anatomical variations and individual characteristics. The deep learning model learns to maintain patient-specific features while achieving normalization, allowing each region of the image to retain its unique structural properties rather than being forced into a generic average anatomy.
Solution Approach 2:
The template is no longer static but dynamically adapted to each patient's anatomy through the deep learning process. The GAN generates templates that are customized to individual patients based on their specific functional image characteristics, making the normalization process adaptive rather than rigidly standardized.
Data Source
AI summary
A device for spatially normalizing a medical image includes: an adaptive template generation unit configured such that when a plurality of functional medical images are input to a deep learning architecture, the adaptive template generation unit generates, based on prestored learning data, an adaptive template for spatially normalizing the plurality of functional medical images; a learning unit configured to learn by repeating a process of generating an image from an input functional medical image of a user based on the adaptive template through a generative adversarial network (GAN) and determine authenticity of the generated image; and a spatial normalization unit configured to provide the functional medical image of the user which is spatially normalized based on results of the learning.