Liver MRI LI-RADS Lexicon for Vision-Language Model Explanation Generation (v2.0)
Abstract
A structured terminology lexicon for LI-RADS major feature description in multiphase liver MRI, developed to guide lexicon-constrained explanation generation by vision-language models (VLMs). The lexicon defines standardized preferred terms, synonyms, absence phrases, and avoid terms for nine LI-RADS radiological features: arterial phase hyperenhancement (APHE), delayed capsule appearance, delayed washout, portal venous capsule appearance, portal venous washout, diffusion restriction, T2 hyperintensity, T1 native lesion signal, and intralesional fat. Version 2.0 incorporates expert radiologist review: CT-specific terms removed, chemical-shift sequences clarified, washout mechanism corrected, and informal phrases removed across all entries. The lexicon was grounded in LI-RADS v2021 guidelines (American College of Radiology) and supplementary MRI literature, and was validated through a blinded pairwise radiologist evaluation study comparing VLM-generated explanations with and without lexicon constraints. Developed at the Biomedical Imaging Group Rotterdam (BIGR), Erasmus MC, as part of a master's thesis on VLM-based liver lesion characterization (Erasmus MC, 2026).
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.