Skip to main content
. 2024 Oct 16;19:144. doi: 10.1186/s13020-024-01005-w

Fig. 1.

Fig. 1

Overall framework of PrescDRL. A Diagnose and treatment process of doctors and intelligence decision model. B The macro framework of PrescDRL. In this sub-figure, MLP represents multi-layer perception. C Reinforcement learning based prediction module of diagnose and treatment scheme. In this sub-figure, FC denotes a fully connect neural layer, and LSTM denotes a neural network of long short-term memory. D Transition network of patient states. In this sub-figure, S and P denotes the states and prescriptions of patient, respectively. E The prediction module of state transition. In this sub-figure, i and j denote i-th and j-th iterations of the deep reinforcement learning model. s, a and r denote the state of patient, the action and reward of the model, respectively