A Unified Attention U-Net Framework for Cross-Modality Tumor Segmentation in MRI and CT

TL;DR

This paper introduces a unified Attention U-Net model trained on both MRI and CT datasets for tumor segmentation, demonstrating strong cross-modality generalization without modality-specific adjustments.

Contribution

The study is among the first to train a single Attention U-Net on separate MRI and CT datasets simultaneously without domain adaptation or modality-specific encoders.

Findings

Achieved competitive Dice, IoU, and AUC scores on both MRI and CT datasets.

Established a robust baseline for cross-modality tumor segmentation.

Validated the model's generalizability across diverse imaging modalities.

Abstract

This study presents a unified Attention U-Net architecture trained jointly on MRI (BraTS 2021) and CT (LIDC-IDRI) datasets to investigate the generalizability of a single model across diverse imaging modalities and anatomical sites. Our proposed pipeline incorporates modality-harmonized preprocessing, attention-gated skip connections, and a modality-aware Focal Tversky loss function. To the best of our knowledge, this study is among the first to evaluate a single Attention U-Net trained simultaneously on separate MRI (BraTS) and CT (LIDC-IDRI) tumor datasets, without relying on modality-specific encoders or domain adaptation. The unified model demonstrates competitive performance in terms of Dice coefficient, IoU, and AUC on both domains, thereby establishing a robust and reproducible baseline for future research in cross-modality tumor segmentation.