A comparative study of reinforcement learning algorithms applied to medical image registration

MESCHINO, GUSTAVO J.; ISA-JARA, RAMIRO F.; BALLARIN, VIRGINIA L.

IFMBE PROCEEDINGS

Springer

Lugar: Laussane; Año: 2019 vol. 75 p. 1 - 8

1680-0737

Nowadays, several medical procedures depend on the comparisonand combination of images obtained in different modalities (magnetic resonance,computed tomography, positron emission tomography and among so on). Imageregistration is a geometric transformation process to align two or more images. It is necessary to have robust algorithms to find the best parameters of transformation in order to achieve accurate registrations. Reinforcement learning allows for an agent to be trained through direct environment interaction to achieve a goal. In this work, a comparison of the performance of Q-learning and Deep Q with its variants is presented. Brain magnetic resonance images are used in a 2D domain considering rigid deformations. The comparison is based on the reward values, computing the Pearson correlation factor in monomodal registration and mutual information in multimodal registration, obtained during the learning process. It is also considered an error measure between the target parameters and the achieved ones. Finally, a backup memory criterion is proposed to train the Q-Network methods. Experimental results show a successful behaviour in all cases, but the performance is improved when the proposed criterion is applied.