During the diagnostic process, clinicians leverage multimodal information, such as the chief complaint, medical images and laboratory test results. Deep-learning models for aiding diagnosis have yet to meet this requirement of leveraging multimodal information. Here we report a transformer-based representation-learning model as a clinical diagnostic aid that processes multimodal input in a unified manner. Rather than learning modality-specific features, the model leverages embedding layers to convert images and unstructured and structured text into visual tokens and text tokens, and uses bidirectional blocks with intramodal and intermodal attention to learn holistic representations of radiographs, the unstructured chief complaint and clinical history, and structured clinical information such as laboratory test results and patient demographic information. The unified model outperformed an image-only model and non-unified multimodal diagnosis models in the identification of pulmonary disease (by 12% and 9%, respectively) and in the prediction of adverse clinical outcomes in patients with COVID-19 (by 29% and 7%, respectively). Unified multimodal transformer-based models may help streamline the triaging of patients and facilitate the clinical decision-making process.

The personalized titration and optimization of insulin regimens for treatment of type 2 diabetes (T2D) are resource-demanding healthcare tasks. Here we propose a model-based reinforcement learning (RL) framework (called RL-DITR), which learns the optimal insulin regimen by analyzing glycemic state rewards through patient model interactions. When evaluated during the development phase for managing hospitalized patients with T2D, RL-DITR achieved superior insulin titration optimization (mean absolute error (MAE) of 1.10 ± 0.03 U) compared to other deep learning models and standard clinical methods. We performed a stepwise clinical validation of the artificial intelligence system from simulation to deployment, demonstrating better performance in glycemic control in inpatients compared to junior and intermediate-level physicians through quantitative (MAE of 1.18 ± 0.09 U) and qualitative metrics from a blinded review. Additionally, we conducted a single-arm, patient-blinded, proof-of-concept feasibility trial in 16 patients with T2D. The primary outcome was difference in mean daily capillary blood glucose during the trial, which decreased from 11.1 (±3.6) to 8.6 (±2.4) mmol L⁻¹ (P < 0.01), meeting the pre-specified endpoint. No episodes of severe hypoglycemia or hyperglycemia with ketosis occurred. These preliminary results warrant further investigation in larger, more diverse clinical studies. ClinicalTrials.gov registration: NCT05409391.

Background. Deep learning has been widely used for glaucoma diagnosis. However, there is no clinically validated algorithm for glaucoma incidence and progression prediction. This study aims to develop a clinically feasible deep-learning system for predicting and stratifying the risk of glaucoma onset and progression based on color fundus photographs (CFPs), with clinical validation of performance in external population cohorts. Methods. We established data sets of CFPs and visual fields collected from longitudinal cohorts. The mean follow-up duration was 3 to 5 years across the data sets. Artificial intelligence (AI) models were developed to predict future glaucoma incidence and progression based on the CFPs of 17,497 eyes in 9346 patients. The area under the receiver operating characteristic (AUROC) curve, sensitivity, and specificity of the AI models were calculated with reference to the labels provided by experienced ophthalmologists. Incidence and progression of glaucoma were determined based on longitudinal CFP images or visual fields, respectively. Results. The AI model to predict glaucoma incidence achieved an AUROC of 0.90 (0.81–0.99) in the validation set and demonstrated good generalizability, with AUROCs of 0.89 (0.83–0.95) and 0.88 (0.79–0.97) in external test sets 1 and 2, respectively. The AI model to predict glaucoma progression achieved an AUROC of 0.91 (0.88–0.94) in the validation set, and also demonstrated outstanding predictive performance with AUROCs of 0.87 (0.81–0.92) and 0.88 (0.83–0.94) in external test sets 1 and 2, respectively. Conclusion. Our study demonstrates the feasibility of deep-learning algorithms in the early detection and prediction of glaucoma progression. FUNDING. National Natural Science Foundation of China (NSFC); the High-level Hospital Construction Project, Zhongshan Ophthalmic Center, Sun Yat-sen University; the Science and Technology Program of Guangzhou, China (2021), the Science and Technology Development Fund (FDCT) of Macau, and FDCT-NSFC.