_version_ 1866913799565475840
author Wang, Meng
Lin, Tian
Lin, Aidi
Yu, Kai
Peng, Yuanyuan
Wang, Lianyu
Chen, Cheng
Zou, Ke
Liang, Huiyu
Chen, Man
Yao, Xue
Zhang, Meiqin
Huang, Binwei
Zheng, Chaoxin
Zhang, Peixin
Chen, Wei
Luo, Yilong
Chen, Yifan
Xia, Honghe
Shi, Tingkun
Zhang, Qi
Guo, Jinming
Chen, Xiaolin
Wang, Jingcheng
Tham, Yih Chung
Liu, Dianbo
Wong, Wendy
Thakur, Sahil
Fenner, Beau
Fang, Danqi
Liu, Siying
Liu, Qingyun
Huang, Yuqiang
Zeng, Hongqiang
Meng, Yanda
Zhou, Yukun
Jiang, Zehua
Qiu, Minghui
Zhang, Changqing
Chen, Xinjian
Wang, Sophia Y.
Lee, Cecilia S.
Sobrin, Lucia
Cheung, Carol Y
Pang, Chi Pui
Keane, Pearse A.
Cheng, Ching-Yu
Chen, Haoyu
Fu, Huazhu
author_facet Wang, Meng
Lin, Tian
Lin, Aidi
Yu, Kai
Peng, Yuanyuan
Wang, Lianyu
Chen, Cheng
Zou, Ke
Liang, Huiyu
Chen, Man
Yao, Xue
Zhang, Meiqin
Huang, Binwei
Zheng, Chaoxin
Zhang, Peixin
Chen, Wei
Luo, Yilong
Chen, Yifan
Xia, Honghe
Shi, Tingkun
Zhang, Qi
Guo, Jinming
Chen, Xiaolin
Wang, Jingcheng
Tham, Yih Chung
Liu, Dianbo
Wong, Wendy
Thakur, Sahil
Fenner, Beau
Fang, Danqi
Liu, Siying
Liu, Qingyun
Huang, Yuqiang
Zeng, Hongqiang
Meng, Yanda
Zhou, Yukun
Jiang, Zehua
Qiu, Minghui
Zhang, Changqing
Chen, Xinjian
Wang, Sophia Y.
Lee, Cecilia S.
Sobrin, Lucia
Cheung, Carol Y
Pang, Chi Pui
Keane, Pearse A.
Cheng, Ching-Yu
Chen, Haoyu
Fu, Huazhu
contents Previous foundation models for fundus images were pre-trained with limited disease categories and knowledge base. Here we introduce a knowledge-rich vision-language model (RetiZero) that leverages knowledge from more than 400 fundus diseases. For RetiZero's pretraining, we compiled 341,896 fundus images paired with texts, sourced from public datasets, ophthalmic literature, and online resources, encompassing a diverse range of diseases across multiple ethnicities and countries. RetiZero exhibits remarkable performance in several downstream tasks, including zero-shot disease recognition, image-to-image retrieval, AI-assisted clinical diagnosis,few-shot fine-tuning, and internal- and cross-domain disease identification. In zero-shot scenarios, RetiZero achieves Top-5 accuracies of 0.843 for 15 diseases and 0.756 for 52 diseases. For image retrieval, it achieves Top-5 scores of 0.950 and 0.886 for the same sets, respectively. AI-assisted clinical diagnosis results show that RetiZero's Top-3 zero-shot performance surpasses the average of 19 ophthalmologists from Singapore, China, and the United States. RetiZero substantially enhances clinicians' accuracy in diagnosing fundus diseases, in particularly rare ones. These findings underscore the value of integrating the RetiZero into clinical settings, where various fundus diseases are encountered.
format Preprint
id arxiv_https___arxiv_org_abs_2406_09317
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Enhancing Diagnostic Accuracy in Rare and Common Fundus Diseases with a Knowledge-Rich Vision-Language Model
Wang, Meng
Lin, Tian
Lin, Aidi
Yu, Kai
Peng, Yuanyuan
Wang, Lianyu
Chen, Cheng
Zou, Ke
Liang, Huiyu
Chen, Man
Yao, Xue
Zhang, Meiqin
Huang, Binwei
Zheng, Chaoxin
Zhang, Peixin
Chen, Wei
Luo, Yilong
Chen, Yifan
Xia, Honghe
Shi, Tingkun
Zhang, Qi
Guo, Jinming
Chen, Xiaolin
Wang, Jingcheng
Tham, Yih Chung
Liu, Dianbo
Wong, Wendy
Thakur, Sahil
Fenner, Beau
Fang, Danqi
Liu, Siying
Liu, Qingyun
Huang, Yuqiang
Zeng, Hongqiang
Meng, Yanda
Zhou, Yukun
Jiang, Zehua
Qiu, Minghui
Zhang, Changqing
Chen, Xinjian
Wang, Sophia Y.
Lee, Cecilia S.
Sobrin, Lucia
Cheung, Carol Y
Pang, Chi Pui
Keane, Pearse A.
Cheng, Ching-Yu
Chen, Haoyu
Fu, Huazhu
Image and Video Processing
Computer Vision and Pattern Recognition
Previous foundation models for fundus images were pre-trained with limited disease categories and knowledge base. Here we introduce a knowledge-rich vision-language model (RetiZero) that leverages knowledge from more than 400 fundus diseases. For RetiZero's pretraining, we compiled 341,896 fundus images paired with texts, sourced from public datasets, ophthalmic literature, and online resources, encompassing a diverse range of diseases across multiple ethnicities and countries. RetiZero exhibits remarkable performance in several downstream tasks, including zero-shot disease recognition, image-to-image retrieval, AI-assisted clinical diagnosis,few-shot fine-tuning, and internal- and cross-domain disease identification. In zero-shot scenarios, RetiZero achieves Top-5 accuracies of 0.843 for 15 diseases and 0.756 for 52 diseases. For image retrieval, it achieves Top-5 scores of 0.950 and 0.886 for the same sets, respectively. AI-assisted clinical diagnosis results show that RetiZero's Top-3 zero-shot performance surpasses the average of 19 ophthalmologists from Singapore, China, and the United States. RetiZero substantially enhances clinicians' accuracy in diagnosing fundus diseases, in particularly rare ones. These findings underscore the value of integrating the RetiZero into clinical settings, where various fundus diseases are encountered.
title Enhancing Diagnostic Accuracy in Rare and Common Fundus Diseases with a Knowledge-Rich Vision-Language Model
topic Image and Video Processing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2406.09317