ABSTRACT

In many medical imaging applications, we need to segment one object of interest. The techniques used for segmentation vary depending on the particular situation and the specifications of the problem at hand. This chapter introduces a new multistage image segmentation system based on reinforcement learning (RL). In this system, the RL agent takes specific actions, such as changing the tasks parameters, to modify the quality of the segmented image. The approach starts with a limited number of training samples and improves its performance in the course of time. It contains an offline mode, where

the reinforcement learning agent uses some images and manually segmented versions of these images to provide the segmentation agent with basic information about the application domain. The reinforcement agent is provided with reward and punishment to explore and exploit the solution space. After using this mode, the agent can choose the appropriate parameter values for different processing tasks on the basis of its accumulated knowledge. The online mode consequently guarantees that the system is continuously training. By using these two learning modes, the RL agent allows us to recognize the parameters for the entire segmentation process. The results on transrectal ultrasound (TRUS) images demonstrate the potential of this approach in the field of medical image segmentation.