LUMEN: Longitudinal Multi-Modal Radiology Model for Prognosis and Diagnosis

TL;DR

LUMEN is a novel training framework for vision-language models that enhances longitudinal chest X-ray interpretation, enabling improved diagnosis and prognosis through multi-image and multi-task instruction fine-tuning.

Contribution

The paper introduces LUMEN, a new training framework optimized for longitudinal radiology data, incorporating multi-image and multi-task instruction fine-tuning for better prognostic and diagnostic performance.

Findings

Significant improvements in diagnostic VQA tasks.

Demonstrated potential for accurate prognostic capabilities.

Enhanced interpretability of longitudinal radiological data.

Abstract

Large vision-language models (VLMs) have evolved from general-purpose applications to specialized use cases such as in the clinical domain, demonstrating potential for decision support in radiology. One promising application is assisting radiologists in decision-making by the analysis of radiology imaging data such as chest X-rays (CXR) via a visual and natural language question-answering (VQA) interface. When longitudinal imaging is available, radiologists analyze temporal changes, which are essential for accurate diagnosis and prognosis. The manual longitudinal analysis is a time-consuming process, motivating the development of a training framework that can provide prognostic capabilities. We introduce a novel training framework LUMEN, that is optimized for longitudinal CXR interpretation, leveraging multi-image and multi-task instruction fine-tuning to enhance prognostic and diagnostic performance. We conduct experiments on the publicly available MIMIC-CXR and its associated Medical-Diff-VQA datasets. We further formulate and construct a novel instruction-following dataset incorporating longitudinal studies, enabling the development of a prognostic VQA task. Our method demonstrates significant improvements over baseline models in diagnostic VQA tasks, and more importantly, shows promising potential for prognostic capabilities. These results underscore the value of well-designed, instruction-tuned VLMs in enabling more accurate and clinically meaningful radiological interpretation of longitudinal radiological imaging data.