Abstract
Most diabetic retinopathy (DR) research focuses on diagnosis, but less attention is given to helping patients to adjust to their new situation. This paper presents a Retrieval-Augmented Generation (RAG) system that gives personalized dietary support, adapting to the patient DR severity stage. The system retrieves recipes from a vector database filtered for allergies and dietary needs, then re-ranked by cuisine preference. We test the system on 15 diabetic patients using two small language models, comparing RAG against a zero-shot baseline.
A case study on the most severe patients further shows how the system behaves in high-risk scenarios.
Results show RAG responses use more diabetes-relevant vocabulary and provide more detailed answers than the baseline. These findings suggest that grounding LLM-based dietary advice in real recipes can help make personalized nutrition support more accessible, especially in low-resource contexts where dietitian access is limited
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MIRASOL_055.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=komVaeZmXV
BibTex
@InProceedings{SelMal_Eyes_MICCAISAT2026,
author = { Selma, Malak Ben AND Morales-Garzón, Andrea AND Martin-Bautista, Maria J.},
title = { { Eyes on diet: a low-resource framework connecting diabetic retinopathy screening to regional nutritional guidance } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17265},
month = {pending},
page = {pending}
}
