Multi-omics MSI Prediction System Using CT and MRI Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting microsatellite instability in colorectal cancer, such as MSI-H/dMMR status, are invasive, time-consuming, and economically costly, with low accuracy due to relying on one-dimensional radiomics or pathomics analysis and requiring surgical specimens for MSI prediction, which delays patient treatment.
Innovation Solution
A system and method that combines pre-treatment enhanced CT and multi-modal MRI images with pathological whole slide image (PWSI) depth features and omics features, using region growing image segmentation and fully-connected neural network algorithms for assisted delineation and feature extraction, along with clinical data analysis to generate radiomics and pathomics signatures for comprehensive MSI prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gene detection or immunohistochemical staining is conducted on colonoscopy biopsy specimens or postoperative large pathological specimens, then MSI-H/dMMR status can be detected, but the detection process increases the traumatic property, time cost, and economic cost
Solution Approach 1:
The patent uses imaging data (CT, MRI, endoscopic images) as non-invasive copies or proxies of the actual tissue characteristics to predict MSI-H/dMMR status, eliminating the need for invasive biopsy sampling while maintaining diagnostic accuracy through radiomics and pathomics feature extraction
Solution Approach 2:
The patent replaces the mechanical/invasive biopsy sampling process with non-invasive imaging-based prediction systems that use computational algorithms to extract diagnostic information from external images, substituting physical tissue extraction with digital image analysis
2Measurement precision
If gene detection or immunohistochemical staining is conducted on postoperative large pathological specimens, then MSI-H/dMMR status can be detected, but the detection process has timeliness lag
Solution Approach 1:
The patent performs MSI-H/dMMR prediction using imaging data obtained before treatment begins, allowing clinicians to make informed decisions about neoadjuvant therapy or surgical planning in advance, rather than waiting for postoperative pathological analysis
Solution Approach 2:
The patent uses pre-treatment imaging data as proxies to predict molecular characteristics, enabling early detection and decision-making without waiting for the time-consuming postoperative pathological analysis process
3Device complexity
If one-dimensional radiomics or pathomics analysis is used for MSI prediction, then the prediction process is simpler, but the prediction accuracy is low
Solution Approach 1:
The patent merges radiomics analysis (from CT/MRI images) with pathomics analysis (from endoscopic images) to create a multi-omics prediction system that combines complementary information from different imaging modalities, thereby improving prediction accuracy while maintaining systematic integration
Solution Approach 2:
The patent creates a composite prediction model that integrates multiple types of imaging data and feature extraction methods, combining the strengths of different imaging modalities to achieve superior prediction performance compared to single-modality approaches
Data Source
AI summary
A system for predicting microsatellite instability and a construction method thereof, a terminal device and a medium are provided. Target image information, pathological specimen information and clinical data information of a user to be predicted are acquired by an acquisition module in the system for predicting microsatellite instability; a radiomics signature is generated according to the target image information, and a pathomics signature is generated according to the pathological specimen information, by a signature generation module, based on a pre-trained MSI-H/dMMR multi-omics signature model; and MSI-H/dMMR prediction results are generated according to the radiomics signature, the pathomics signature and the clinical data information, by a prediction generation module, based on a pre-trained MSI-H/dMMR prediction model.