Automatic SPECT Image Reorientation via Deep Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing cardiac SPECT images require manual reorientation of the left ventricle to a standard view, introducing random errors and reducing analysis accuracy due to subjective operation, which is time-consuming.
Innovation Solution
An automatic reorientation method using a deep learning network, comprising a feature extraction network and a spatial transformer network, extracts rigid registration parameters to rotate SPECT three-dimensional reconstructed images to a standard view, improving accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual reorientation of left ventricle to standard view is performed, then clinical analysis can be conducted, but random errors are introduced and analysis accuracy is reduced
Solution Approach 1:
The patent replaces the manual mechanical reorientation operation with an automated deep learning-based image processing system. The network automatically performs the reorientation from conventional view to standard view, eliminating subjective manual manipulation and its associated random errors, thereby improving measurement precision while reducing operational complexity
Solution Approach 2:
The system enables self-service by allowing the SPECT image to be automatically reoriented through the deep learning network without requiring manual intervention. The network independently processes the image transformation, making the system self-sufficient and eliminating the need for operator involvement in the reorientation process
2Productivity
If manual reorientation operation is performed, then standard view is obtained, but time consumption increases
Solution Approach 1:
The patent substitutes the time-consuming manual reorientation process with an automated deep learning network that processes images rapidly. This mechanical substitution eliminates the slow manual operation while maintaining or improving the quality of the reorientation, thereby increasing productivity and reducing time loss
Solution Approach 2:
The system performs preliminary automated processing by pre-training the deep learning network on labeled datasets, enabling it to quickly perform reorientation operations without requiring manual preparation or adjustment during actual use, thus reducing operational time
3Measurement precision
If deep learning network is used for automatic reorientation, then reorientation accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the deep learning network into distinct functional modules: a feature extraction network and a spatial transformer network. This segmentation allows each module to perform a specific function (feature extraction and spatial transformation respectively), making the overall complex system more manageable and interpretable while maintaining high reorientation accuracy
Solution Approach 2:
The patent introduces a spatial transformer network as an intermediary between feature extraction and final image transformation. This intermediary component processes the extracted features and generates the appropriate transformation parameters, bridging the gap between feature representation and spatial reorientation while managing system complexity
Data Source
AI summary
Disclosed is an automatic reorientation method from an SPECT three-dimensional reconstructed image to a standard view, wherein a rigid registration parameter P between a SPECT three-dimensional reconstructed image A and a standard SPECT image R is extracted by using a rigid registration algorithm to form a mapping database of A and P; features of the image A are extracted by using a three-layer convolution module, and are converted into a 6-dimensional feature vector T after three times of full connection, and T is applied to A through a spatial transformer network to form an orientation result predicted by the network, thus establishing the automatic reorientation model of the SPECT three-dimensional reconstructed image. The SPECT three-dimensional reconstructed image to be orientated is taken as an input. A standard view can be obtained by using the automatic reorientation model of the SPECT three-dimensional reconstructed image for automatic turning.