A Multi-Stage Fine-Tuning and Ensembling Strategy for Pancreatic Tumor Segmentation in Diagnostic and Therapeutic MRI

2025-08-29

AI Summary

Pancreatic ductal adenocarcinoma (PDAC) segmentation in T1-weighted (Task 1) and T2-weighted (Task 2) MRI is challenging due to low tissue contrast and scarce expert annotations. To address data scarcity and domain shift, we propose a multi-stage cascaded pretraining and metric-aware heterogeneous ensemble method built upon nnU-Net: (i) CT-to-MRI transfer learning mitigates limited MRI annotation; (ii) expert models are trained with modality-specific augmentation intensities; and (iii) ensemble weights dynamically balance volumetric overlap (Dice) and boundary accuracy (HD95, MASD). Under five-fold cross-validation, our method achieves Dice scores of 0.661 (Task 1) and 0.523 (Task 2), with MASD = 5.46 mm and HD95 = 17.33 mm—outperforming state-of-the-art approaches. The core contribution is a clinically motivated, robust multimodal MRI segmentation framework leveraging heterogeneous model ensembling guided by complementary geometric and volumetric metrics.

Technology Category

• Machine Learning: Ensemble Methods
• Computer Vision: Segmentation
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

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

Abstract

Automated segmentation of Pancreatic Ductal Adenocarcinoma (PDAC) from MRI is critical for clinical workflows but is hindered by poor tumor-tissue contrast and a scarcity of annotated data. This paper details our submission to the PANTHER challenge, addressing both diagnostic T1-weighted (Task 1) and therapeutic T2-weighted (Task 2) segmentation. Our approach is built upon the nnU-Net framework and leverages a deep, multi-stage cascaded pre-training strategy, starting from a general anatomical foundation model and sequentially fine-tuning on CT pancreatic lesion datasets and the target MRI modalities. Through extensive five-fold cross-validation, we systematically evaluated data augmentation schemes and training schedules. Our analysis revealed a critical trade-off, where aggressive data augmentation produced the highest volumetric accuracy, while default augmentations yielded superior boundary precision (achieving a state-of-the-art MASD of 5.46 mm and HD95 of 17.33 mm for Task 1). For our final submission, we exploited this finding by constructing custom, heterogeneous ensembles of specialist models, essentially creating a mix of experts. This metric-aware ensembling strategy proved highly effective, achieving a top cross-validation Tumor Dice score of 0.661 for Task 1 and 0.523 for Task 2. Our work presents a robust methodology for developing specialized, high-performance models in the context of limited data and complex medical imaging tasks (Team MIC-DKFZ).

Problem

• Automated segmentation of pancreatic tumors from MRI
• Overcoming poor tumor-tissue contrast in medical imaging
• Addressing scarcity of annotated medical imaging data

Innovation

• Multi-stage cascaded pre-training strategy
• Custom heterogeneous ensembles of models
• Metric-aware ensembling for optimal performance