Registration accuracy assessment on noisy neuroimages

Validation and accuracy assessment are the main bottlenecks preventing the adoption of many medical image processing algorithms in the clinical practice. In the classical approach, a-posteriori analysis is performed based on some predefined objective metrics. In this paper, a different approach based on Petri Nets is proposed. The basic idea consists in predicting the accuracy that will result from a given processing on a given type of data based on the identification and characterization of the sources of inaccuracy intervening along the whole chain. Here it is proposed a proof of concept in the specific case of noisy Magnetic Resonance image registration. Magnetic Resonance images are normally corrupted by random noise from the measurement process complicating the automatic feature extraction and analysis of clinical data. The accurate registration of images observed in additive noise is a challenging task. The noise can increase the number of misregistered regions, and decrease the accuracy of subpixel registration. A Petri Net is built after the detection of the possible sources of inaccuracy, ranging from the images noise to the registration parameters adopted, and the evaluation of their respective impact on the estimation of the deformation field. A training set of five different synthetic volumes is used. Afterward, validation is performed on a different set of five synthetic volumes by comparing the estimated inaccuracy with the posterior measurements according to a set of predefined metrics. Results show that the proposed model provides a good prediction performance. An extended set of clinical data will allow the complete characterization of the system for the considered task.