LoGSAM: Parameter-Efficient Cross-Modal Grounding for MRI Segmentation

Islam Bhuiyan MR, Bhat S, Qahqaie M, Nguyen TT, Perez-Toro PA, Arias-Vergara T, Maier A (2026)

Publication Language: English

Publication Status: Submitted

Publication Type: Unpublished / Preprint

Future Publication Type: Conference contribution

Publication year: 2026

City/Town: https://arxiv.org/abs/2603.17576

URI: https://arxiv.org/abs/2603.17576

DOI: https://arxiv.org/abs/2603.17576

Open Access Link: https://arxiv.org/abs/2603.17576

Abstract

Precise localization and delineation of brain tumors using Magnetic Resonance Imaging (MRI) are essential for planning therapy and guiding surgical decisions. However, most existing approaches rely on task-specific supervised models and are constrained by the limited availability of annotated data. To address this, we propose LoGSAM, a parameter-efficient, detection-driven framework that transforms radiologist dictation into text prompts for foundation-model-based localization and segmentation. Radiologist speech is first transcribed and translated using a pretrained Whisper ASR model, followed by negation-aware clinical NLP to extract tumor-specific textual prompts. These prompts guide text-conditioned tumor localization via a LoRA-adapted vision-language detection model, Grounding DINO (GDINO). The LoRA adaptation updates using 5% of the model parameters, thereby enabling computationally efficient domain adaptation while preserving pretrained cross-modal knowledge. The predicted bounding boxes are used as prompts for MedSAM to generate pixel-level tumor masks without any additional fine-tuning. Conditioning the frozen MedSAM on LoGSAM-derived priors yields a state-of-the-art dice score of 80.32% on BRISC 2025. In addition, we evaluate the full pipeline using German dictations from a board-certified radiologist on 12 unseen MRI scans, achieving 91.7% case-level accuracy. These results highlight the feasibility of constructing a modular, speech-to-segmentation pipeline by intelligently leveraging pretrained foundation models with minimal parameter updates.

Authors with CRIS profile

Involved external institutions

Universidad de Antioquía (UDEA) Colombia (CO) King’s College London United Kingdom (GB)

How to cite

APA:

Islam Bhuiyan, M.R., Bhat, S., Qahqaie, M., Nguyen, T.-T., Perez-Toro, P.A., Arias-Vergara, T., & Maier, A. (2026). LoGSAM: Parameter-Efficient Cross-Modal Grounding for MRI Segmentation. (Unpublished, Submitted).

MLA:

Islam Bhuiyan, Mohammad Robaitul, et al. LoGSAM: Parameter-Efficient Cross-Modal Grounding for MRI Segmentation. Unpublished, Submitted. 2026.

BibTeX: Download