Partial Information Decomposition as a Multi-Contrast 3D MRI Selection Strategy for Resource-Constrained Deep Neural Network Training in Brain Tumor Segmentation
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
The authors looked at how to pick the best two types of MRI images out of four to use for identifying brain tumors, without needing to process all the images and slow things down. They used a method called Partial Information Decomposition to measure how much unique and shared information each pair of images had about the tumor. Their method chose the T1c and T2-FLAIR images as the best pair for training their tumor segmentation models. Tests showed this pair worked nearly as well as using all four images. This approach helps decide which image types to use before building complex MRI analysis models.
Authors
Agamdeep Chopra, Mehmet Kurt
Abstract
Multi-contrast 3D MRI segmentation can be computationally demanding when all available sequences are used. We evaluate a pre-training Partial Information Decomposition framework that ranks input pairs according to their redundant, unique, and synergistic information about regional tumor burden and selects the highest-ranked pair for downstream training. Applied to T1n, T1c, T2w, and T2-FLAIR MRI, the framework selected T1c+T2-FLAIR. We then trained eleven architecturally identical lightweight 3D U-Nets using different input configurations. On an independent test cohort, T1c+T2-FLAIR was the strongest two-input configuration and ranked second overall in mean Dice (0.676 versus 0.687 for all four inputs). Independent Shapley analysis on the full-input model also identified T2-FLAIR and T1c as the most influential inputs and their pairwise interaction as the strongest. These findings demonstrate the practical value of PID based pre-training selection for identifying compact, informative MRI input sets before costly 3D model development.