LUMEN: Longitudinal Multi-Modal Radiology Model for Prognosis and Diagnosis
2026-02-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors developed a new training system called LUMEN to help large vision-language models better analyze chest X-rays taken over time. This system helps these models answer questions about both current and changing features in the images, which is important for doctors to make diagnoses and predictions. They tested their approach using public datasets and created new instructions for the model to handle sequences of images. Their results show improved accuracy in answering diagnostic questions and potential for predicting patient outcomes, which could support radiologists in clinical work.
vision-language modelsradiologychest X-rayslongitudinal imagingquestion answeringinstruction fine-tuningprognosisdiagnosisMIMIC-CXRVQA
Authors
Zhifan Jiang, Dong Yang, Vishwesh Nath, Abhijeet Parida, Nishad P. Kulkarni, Ziyue Xu, Daguang Xu, Syed Muhammad Anwar, Holger R. Roth, Marius George Linguraru
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.