Article
作者: Tham, Yih-Chung ; Luk, Andrea O Y ; Corsino, Leonor ; Guan, Zhouyu ; Raman, Rajiv ; Cheng, Di ; Bascaran, Covadonga ; Jia, Weiping ; Sheng, Bin ; Hu, Tingting ; Tan, Hong Chang ; Liu, Ruhan ; Huang, Shan ; Yu, Shujie ; Gu, Jing ; Bao, Yuqian ; Ruan, Lei ; Sun, Lina ; Li, Huating ; Qin, Yiming ; Wong, Tien Yin ; Chotcomwongse, Peranut ; Zhang, Xiao ; Lim, Lee-Ling ; Cheng, Ching-Yu ; Yin, Jun ; Huang, Gengyou ; Wu, Liang ; Ma, Weizhi ; Zhang, Chenxi ; Sun, Xiaodong ; Bee, Yong Mong ; Echouffo-Tcheugui, Justin B ; Nguyen, Truong X ; Yang, Xiaolong ; Dai, Ling ; Sivaprasad, Sobha ; Cleland, Charles R ; Lu, Yuwei ; Shaw, Jonathan E ; Ruamviboonsuk, Paisan ; Liew, Zhong Hong ; Wang, Xiangning ; Goh, Jocelyn Hui Lin ; Ekinci, Elif I ; Wang, Zheyuan ; Chen, Tingli ; Mok, Shao Feng ; Zheng, Yingfeng ; Simó, Rafael ; Jiang, Nan ; Yang, Xiaokang ; Ibragimova, Nilufar ; Li, Tingyao ; Wang, Jing ; Li, Jiajia ; Huang, Wenyong ; Guo, Minyi ; Srinivasan, Ramyaa ; Wang, Ya Xing ; Wu, Qiang ; Chen, Li Jia ; Cheung, Carol Y ; Shu, Jia ; Hee, Nicholas Ken Yoong ; Tien, Carolyn Shan-Yeu ; Quek, Ten Cheer ; Wu, Chan ; Wan, Jing ; Guan, Xinping ; Abdullah, Adina ; Normatova, Nargiza ; Zhong, Pingting ; Malek, Rachid ; Chan, Juliana C N ; Liu, Dan ; Du, Wei ; Dai, Rongping ; Zhou, Yan ; Deng, Chenxin ; Hamzah, Haslina ; Keane, Pearse A ; Hu, Xiaoyan ; Yang, Dawei ; Tan, Gavin Siew Wei ; Jenkins, Alicia ; Meng, Ziyao ; Koh, Sky Wei Chee ; Szeto, Simon K H ; Kao, Shih Ling ; Zhang, Cuntai ; Lim, Cynthia Ciwei ; Lim, Amanda Yuan Ling
AbstractPrimary diabetes care and diabetic retinopathy (DR) screening persist as major public health challenges due to a shortage of trained primary care physicians (PCPs), particularly in low-resource settings. Here, to bridge the gaps, we developed an integrated image–language system (DeepDR-LLM), combining a large language model (LLM module) and image-based deep learning (DeepDR-Transformer), to provide individualized diabetes management recommendations to PCPs. In a retrospective evaluation, the LLM module demonstrated comparable performance to PCPs and endocrinology residents when tested in English and outperformed PCPs and had comparable performance to endocrinology residents in Chinese. For identifying referable DR, the average PCP’s accuracy was 81.0% unassisted and 92.3% assisted by DeepDR-Transformer. Furthermore, we performed a single-center real-world prospective study, deploying DeepDR-LLM. We compared diabetes management adherence of patients under the unassisted PCP arm (n = 397) with those under the PCP+DeepDR-LLM arm (n = 372). Patients with newly diagnosed diabetes in the PCP+DeepDR-LLM arm showed better self-management behaviors throughout follow-up (P < 0.05). For patients with referral DR, those in the PCP+DeepDR-LLM arm were more likely to adhere to DR referrals (P < 0.01). Additionally, DeepDR-LLM deployment improved the quality and empathy level of management recommendations. Given its multifaceted performance, DeepDR-LLM holds promise as a digital solution for enhancing primary diabetes care and DR screening.