PEMMA: Parameter-Efficient Multi-Modal Adaptation for Medical Image Segmentation

TL;DR

PEMMA introduces a parameter-efficient framework that enables transformer-based medical image segmentation models to adapt from CT scans to also incorporate PET scans, achieving high performance with minimal additional parameters.

Contribution

The paper presents a novel low-rank adaptation method for transformers that allows flexible multi-modal integration in medical imaging with minimal parameter increase.

Findings

Achieves comparable performance to early fusion with only 8% of trainable parameters.

Provides a +28% improvement in dice score on PET scans when trained on a single modality.

Enables updating the model with only one modality without forgetting the other.

Abstract

Imaging modalities such as Computed Tomography (CT) and Positron Emission Tomography (PET) are key in cancer detection, inspiring Deep Neural Networks (DNN) models that merge these scans for tumor segmentation. When both CT and PET scans are available, it is common to combine them as two channels of the input to the segmentation model. However, this method requires both scan types during training and inference, posing a challenge due to the limited availability of PET scans, thereby sometimes limiting the process to CT scans only. Hence, there is a need to develop a flexible DNN architecture that can be trained/updated using only CT scans but can effectively utilize PET scans when they become available. In this work, we propose a parameter-efficient multi-modal adaptation (PEMMA) framework for lightweight upgrading of a transformer-based segmentation model trained only on CT scans to also incorporate PET scans. The benefits of the proposed approach are two-fold. Firstly, we leverage the inherent modularity of the transformer architecture and perform low-rank adaptation (LoRA) of the attention weights to achieve parameter-efficient adaptation. Secondly, since the PEMMA framework attempts to minimize cross modal entanglement, it is possible to subsequently update the combined model using only one modality, without causing catastrophic forgetting of the other modality. Our proposed method achieves comparable results with the performance of early fusion techniques with just 8% of the trainable parameters, especially with a remarkable +28% improvement on the average dice score on PET scans when trained on a single modality.