Integrated image-based deep learning and language models for primary diabetes care.
Jiajia LiZhouyu GuanJing WangCarol Y CheungYingfeng ZhengLee-Ling LimCynthia Ciwei LimPaisan RuamviboonsukRajiv RamanLeonor CorsinoJustin B Echouffo-TcheuguiAndrea O Y LukGuy L J ChenXiaodong SunHaslina HamzahQiang WuXiang-Ning WangRuhan LiuYa-Xing WangTingli ChenXiao ZhangXiaolong YangJun YinJing WanWei DuTen Cheer QuekJocelyn Hui Lin GohDawei YangXiaoyan HuTruong X NguyenSimon K H SzetoPeranut ChotcomwongseRachid MalekNargiza NormatovaNilufar IbragimovaRamyaa SrinivasanPingting ZhongWenyong HuangChenxin DengLei RuanCuntai ZhangChenxi ZhangYan ZhouChan WuRongping DaiSky Wei Chee KohAdina AbdullahNicholas Ken Yoong HeeHong Chang TanZhong Hong LiewCarolyn Shan-Yeu TienShih Ling KaoAmanda Yuan Ling LimShao Feng MokLina SunJing GuLiang WuTingyao LiDi ChengZheyuan WangYiming QinLing DaiZiyao MengJia ShuYuwei LuNan JiangTingting HuShan HuangGengyou HuangShujie YuDan LiuWeizhi MaMinyi GuoXinping GuanXiaokang YangCovadonga BascaranCharles R ClelandYuqian BaoElif I EkinciAndrzej S JanuszewskiJuliana Chung Ngor ChanYong Mong BeeSobha SivaprasadJonathan E ShawRafael SimòPearse Andrew KeaneChing-Yu ChengGavin Siew Wei TanWeiping JiaYih-Chung ThamHuating LiBin ShengTien-Yin WongPublished in: Nature medicine (2024)
Primary 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.
Keyphrases
- deep learning
- primary care
- newly diagnosed
- editorial comment
- diabetic retinopathy
- public health
- type diabetes
- cardiovascular disease
- glycemic control
- autism spectrum disorder
- artificial intelligence
- end stage renal disease
- machine learning
- chronic kidney disease
- prognostic factors
- metabolic syndrome
- skeletal muscle
- resistance training