LoRA-based methods on Unet for transfer learning in Subarachnoid Hematoma Segmentation

TL;DR

This study addresses automated aneurysmal SAH segmentation under limited labeled data and cross-institution generalization by examining transfer learning from traumatic brain injury hematomas using Unet. It introduces parameter-efficient LoRA/DoRA methods, including a novel CP-LoRA with Canonical Polyadic decomposition and DoRA variants, and benchmarks them against standard fine-tuning on a multi-view Unet. Results show LoRA/DoRA methods consistently outperform conventional fine-tuning, with DoRA-C excelling overall and CP-LoRA achieving similar performance with far fewer trainable parameters; higher ranks also improve performance, highlighting an over-parameterization effect. The work demonstrates feasible cross-hematoma transfer learning for SAH segmentation and points to practical clinical benefits, particularly for larger hemorrhages, while noting the need for larger, multi-institution datasets for robust deployment.

Abstract

Aneurysmal subarachnoid hemorrhage (SAH) is a life-threatening neurological emergency with mortality rates exceeding 30%. Transfer learning from related hematoma types represents a potentially valuable but underexplored approach. Although Unet architectures remain the gold standard for medical image segmentation due to their effectiveness on limited datasets, Low-Rank Adaptation (LoRA) methods for parameter-efficient transfer learning have been rarely applied to convolutional neural networks in medical imaging contexts. We implemented a Unet architecture pre-trained on computed tomography scans from 124 traumatic brain injury patients across multiple institutions, then fine-tuned on 30 aneurysmal SAH patients from the University of Michigan Health System using 3-fold cross-validation. We developed a novel CP-LoRA method based on tensor CP-decomposition and introduced DoRA variants (DoRA-C, convDoRA, CP-DoRA) that decompose weight matrices into magnitude and directional components. We compared these approaches against existing LoRA methods (LoRA-C, convLoRA) and standard fine-tuning strategies across different modules on a multi-view Unet model. LoRA-based methods consistently outperformed standard Unet fine-tuning. Performance varied by hemorrhage volume, with all methods showing improved accuracy for larger volumes. CP-LoRA achieved comparable performance to existing methods while using significantly fewer parameters. Over-parameterization with higher ranks consistently yielded better performance than strictly low-rank adaptations. This study demonstrates that transfer learning between hematoma types is feasible and that LoRA-based methods significantly outperform conventional Unet fine-tuning for aneurysmal SAH segmentation.

Figures (5)

• Figure 1: Two-level multi-view Unet architecture.
• Figure 2: Standard Unet architecture.
• Figure 3: Patient Dice scores of DoRA-C evaluated models on aneurysmal SAH dataset for different choices of rank.
• Figure 4: Segmentation masks produced by different models. LoRA/DoRA methods use rank 64.
• Figure 5: Predicted and annotated SAH volumes from the DoRA-C model with rank 64. The solid line is the fitted linear regression with 95% confidence intervals and the dashed one is the calibration line.