Multi-Modal Reconstruction Network for Faster High-Resolution Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical imaging systems face challenges in efficiently generating reconstructed images from multi-modal data, particularly due to time and resource constraints, processing power limitations, and incompatibility with certain data formats or outdated reconstruction techniques.
Innovation Solution
An artificial neural network is trained to generate high-resolution reconstructed volumes by combining low-resolution volumes from emission data with higher-resolution volumes from different imaging modalities, such as PET and CT scans, without requiring segmentation or registration with additional modality data, using techniques like conjugate gradient and filtered back-projection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional quantitative iterative reconstruction techniques are used to generate reconstructed images from multi-modal data, then measurement precision and manufacturing precision are improved, but loss of time and productivity deteriorate due to time-consuming processing
Solution Approach 1:
The system performs segmentation and registration of anatomical images (CT or MR) in advance before the emission data acquisition. The attenuation map is pre-computed based on the segmented anatomical structures, so that during reconstruction, only the emission data needs to be processed using this pre-prepared attenuation information, significantly reducing reconstruction time while maintaining precision
Solution Approach 2:
The anatomical imaging data is segmented into different tissue types (bone, soft tissue, air) to create a detailed attenuation map. This segmentation allows the system to apply tissue-specific attenuation coefficients during reconstruction, improving precision while the pre-computed nature of this segmentation reduces the time burden during actual reconstruction
2Manufacturing precision
If conventional reconstruction systems are used, then manufacturing precision is improved, but device complexity increases due to requirements for segmentation and registration capabilities
Solution Approach 1:
The complex segmentation and registration operations are performed in advance during a separate preprocessing step, allowing the main reconstruction system to focus solely on the iterative reconstruction algorithm. This separation reduces the operational complexity of the reconstruction system while maintaining high precision through the use of pre-computed attenuation maps
Solution Approach 2:
The attenuation map serves as an intermediary that bridges the anatomical imaging data and the emission data reconstruction. By pre-processing the anatomical data into this intermediate attenuation representation, the system simplifies the reconstruction process while maintaining the ability to incorporate detailed tissue information for high precision
3Productivity
If high-performance computing systems are used to perform segmentation and reconstruction, then manufacturing precision and productivity are improved, but cost increases due to processing power requirements
Solution Approach 1:
The computationally intensive segmentation and registration are performed once in advance, creating reusable attenuation maps that can be applied to multiple emission data sets. This preliminary computation amortizes the high resource cost over multiple reconstructions, improving overall productivity while maintaining precision
Solution Approach 2:
The system performs segmentation at a level of detail that exceeds what is strictly necessary for basic reconstruction, creating comprehensive attenuation maps that can be used for multiple purposes and patient cases. This excessive preliminary action reduces the need for repeated high-resource operations
Data Source
AI summary
A system and method include training of an artificial neural network to generate an output data set, the training based on the plurality of sets of emission data acquired using a first imaging modality and respective ones of data sets acquired using a second imaging modality.