Improving Pre-trained Segmentation Models using Post-Processing

2025-12-16

AI Summary

Pre-trained segmentation models exhibit poor generalizability, high false-positive rates, label swapping, and inter-slice discontinuity when applied to multimodal MRI (mpMRI) of gliomas—key limitations hindering clinical deployment. Method: We propose a retraining-free adaptive post-processing paradigm that jointly integrates morphological constraints, cross-modal consistency verification, and slice-wise temporal smoothing. This is the first clinically informed, lightweight post-processing framework tailored for large pre-trained models. Contribution/Results: The method preserves computational fairness and environmental sustainability while significantly enhancing segmentation robustness and anatomical plausibility. On the BraTS 2025 Challenge, it achieves a +14.9% Dice improvement on the Sub-Saharan African glioma task and +0.9% on the adult glioma task, demonstrating strong clinical applicability and cross-population generalizability.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multimodal Learning
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Gliomas are the most common malignant brain tumors in adults and are among the most lethal. Despite aggressive treatment, the median survival rate is less than 15 months. Accurate multiparametric MRI (mpMRI) tumor segmentation is critical for surgical planning, radiotherapy, and disease monitoring. While deep learning models have improved the accuracy of automated segmentation, large-scale pre-trained models generalize poorly and often underperform, producing systematic errors such as false positives, label swaps, and slice discontinuities in slices. These limitations are further compounded by unequal access to GPU resources and the growing environmental cost of large-scale model training. In this work, we propose adaptive post-processing techniques to refine the quality of glioma segmentations produced by large-scale pretrained models developed for various types of tumors. We demonstrated the techniques in multiple BraTS 2025 segmentation challenge tasks, with the ranking metric improving by 14.9 % for the sub-Saharan Africa challenge and 0.9% for the adult glioma challenge. This approach promotes a shift in brain tumor segmentation research from increasingly complex model architectures to efficient, clinically aligned post-processing strategies that are precise, computationally fair, and sustainable.

Problem

• Improves glioma segmentation accuracy from pre-trained models
• Reduces systematic errors like false positives and discontinuities
• Enhances computational fairness and sustainability in medical imaging

Innovation

• Adaptive post-processing refines glioma segmentation quality
• Techniques improve BraTS challenge metrics significantly
• Shift from complex models to efficient post-processing strategies