Clinical decision support system based on deep reinforcement learning

System introduction

Existing AI-based clinical decision support systems (CDSSs) are knowledge-based and primarily serve diagnostics. On the other hand, AlphaGo defeated the world champion, so that deep reinforcement learning can use process data to discover new Go knowledge or patterns, so as to surpass humans.

Clinical decision support systems can also greatly improve intelligence through deep reinforcement learning, especially in the direction of disease treatment. Harvard & MIT Medical School and others published a paper in Nature Medicine in 2019 on the technical application of reinforcement learning in medical health. Provide guidelines for the use of reinforcement learning for patient treatment decisions, which they hope will speed up the speed with which observation groups inform medical practices in a safe, risk-aware manner.

Our artificial intelligence team has started research and development in this area since 2018, including data construction, algorithm development, software system development, etc. Our product name is KOKOGPT. KOKOGPT is taken from the I Ching and has a precise meaning.

In terms of data construction, he obtained access to the intensive care database MIT, and then localized, developed data alignment software, and dimensionality reduction tools based on causal testing. In terms of algorithm development, treatment strategies for sepsis,

We used the Deep Deterministic Policy Gradient (DDPG) algorithm to surpass the MIT team's DQN algorithm to achieve higher survival rates for patients. At the same time, the model training time is faster and the stability is better. In terms of software system development, data and algorithms are integrated to obtain data from HIS and LIS in real time, and the disease status and treatment suggestions of each patient are regularly output through the visual interface for doctors' reference.

In addition to the treatment strategy of sepsis, we have also developed treatment strategies for respiratory failure, kidney failure, tumor and other diseases with several tertiary hospitals.