Fig. 1
From: Active guidance in ultrasound bladder scanning using reinforcement learning
An overview of the presented method for navigation an ultrasound probe . At each time step, a RL model, specifically a deep Q network, receives the current ultrasound image as d states, along with the corresponding e reward from the b bladder simulation environment. The optimal movement action is selected from the c action space based on the maximum output Q value. The action space encompasses both translation and tilt, where the details for different settings are described in.
