Partial Information Decomposition as a Multi-Contrast 3D MRI Selection Strategy for Resource-Constrained Deep Neural Network Training in Brain Tumor Segmentation

2026-07-16

AI Summary

This study addresses the challenge of high computational cost in multi-contrast 3D MRI brain tumor segmentation under resource-constrained settings. It introduces, for the first time, a partial information decomposition (PID) framework combined with Shapley value analysis to quantify redundant, unique, and synergistic information among T1n, T1c, T2w, and T2-FLAIR sequences—without requiring full model training—enabling efficient selection of an optimal dual-modality combination. Experimental results demonstrate that the selected T1c+T2-FLAIR pair achieves a Dice score of 0.676 on a lightweight 3D U-Net, closely approaching the performance of the full four-modality input (0.687). This approach significantly reduces computational overhead while preserving high segmentation accuracy, confirming its effectiveness and practicality.

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.

Problem

• Multi-contrast MRI
• Brain Tumor Segmentation
• Resource-Constrained Training
• Input Selection
• 3D U-Net

Innovation

• Partial Information Decomposition
• Multi-contrast MRI
• Brain Tumor Segmentation
• Input Selection
• Lightweight 3D U-Net