<ul><li>Deep reinforcement learning algorithms are effective in plane reformatting tasks, particularly for 4D flow MRI.</li><li>A novel technique is introduced in this paper that uses a flexible coordinate system based on the current state for navigation in volumes at any position or orientation.</li><li>The Asynchronous Advantage Actor Critic (A3C) algorithm for reinforcement learning outperforms Deep Q Network (DQN) in the context of plane reformatting for 4D flow MRI.</li><li>Experimental results show improved accuracy in plane reformatting angular and distance errors, with statistically equivalent flow measurements to those determined by an expert, making the method promising for other medical imaging applications.</li></ul>

Adaptive Plane Reformatting for 4D Flow MRI using Deep Reinforcement Learning

Discover more