Optimized glycemic control of type 2 diabetes with reinforcement learning: a proof-of-concept trial

Abstract

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.

Automated Diagnostic Reports from Images of Electrocardiograms at the Point-of-Care

“…First, we used a fine-tuned DistilBERT model to quantify the semantic similarity between reference and model-generated diagnosis statements. 15–17 Next, we employed natural language generation (NLG) metrics conventionally used to evaluate models designed for image captioning or translation tasks to assess the syntactic similarity between statements. These included ROUGE, BLEU, and METEOR, ranging from 0 to 1, and CIDEr, ranging from 0 to 5, with higher values indicating better overlap between reference and generated text.…”

“…To evaluate the semantic similarity between original and model-generated diagnosis statements, we fine-tuned a lightweight DistilBERT model, 15,36 pretrained on a large corpus of electronic health record notes, 16,17 in the same set of cardiologist-confirmed diagnosis statements used to train the vision-text transformer model. The training mirrored the standard approach for training masked language models, with a chunk size of 128 tokens and a masking probability of 0.15.…”

Can Machine Learning Personalize Cardiovascular Therapy in Sepsis?

“…Sepsis resuscitation systems face particular regulatory challenges due to uncertainty over the gold standard doses of IV fluids and vasopressors in each resuscitation scenario for comparison with the system’s recommendations. A recent study of a reinforcement learning system for insulin prescriptions addressed similar problems by using a blinded panel of expert endocrinologists to evaluate the quality of treatment recommendations ( 57 ). The generalization issues highlighted above have led some to advocate that models be recalibrated in each new setting where they are deployed ( 58 , 59 ), but current medical device regulations lack well-defined pathways for approving machine learning systems that update (“retrain”) their prediction engines.…”

Evaluating the Human Safety Net: Observational study of Physician Responses to Unsafe AI Recommendations in high-fidelity Simulation

“…Artificial Intelligence (AI) driven systems are set to take an increasingly prominent supportive role in decision-making including in high-stakes settings 1,2 . While the final decision remains in human hands, understanding how AI recommendations impact their user's behaviour is crucial.…”