A WDLoRA-Based Multimodal Generative Framework for Clinically Guided Corneal Confocal Microscopy Image Synthesis in Diabetic Neuropathy

Xin Zhang, Liangxiu Han, Yue Shi, Yalin Zheng, Uazman Alam, Maryam Ferdousi, Rayaz Malik · 2026

Read A WDLoRA-Based Multimodal Generative Framework for Clinically Guided Corneal Confocal Microscopy Image Synthesis in Diabetic Neuropathy by Xin Zhang, Liangxiu Han, Yue Shi, Yalin Zheng, Uazman Alam, Maryam Ferdousi, Rayaz Malik (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-13693.txt.

Description

Corneal imaging sensitively reveals small-fiber damage in diabetic nerve disease, but auto diagnosis is limited by scarce labeled data and fine-grained variation in corneal nerve shape. Foundation image models excel at natural images yet struggle with medical anatomy. This framework adapts them with Weight-Decomposed Low-Rank Adaptation, splitting size and direction weight updates so the model separately learns nerve shape and stromal contrast. It is conditioned jointly on nerve masks and disease-specific clinical prompts.

A three-part check found state-of-the-art visual quality, with a Frechet Inception Distance of 5.18, and shape soundness, with a shape similarity index of 0.630, beating GAN and standard diffusion baselines. Synthetic images preserved gold-standard clinical biomarkers, matched real patient data in stats, and improved diagnostic accuracy by 2.1 percent. Xin Zhang leads the work with seven co-authors.

Catalog record

CREATORXin Zhang, Liangxiu Han, Yue Shi, Yalin Zheng, Uazman Alam, Maryam Ferdousi, Rayaz Malik
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE68 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-13693

Browse the full library · About the Tower of Babel