Automatic Segmentation of Lateral Ventricles in MRI of Pediatric Patients from 4 to 12 Months

Abstract

The utilization of image processing tools for magnetic resonance image (MRI) processing has made possible carrying out the study of injuries and diseases, as it facilitates the look up of the information contained in the image. The objective of this work is the implementation of a digital image processing methodology to perform the segmentation of lateral ventricles in MRI images of pediatric patients using automatic thresholding techniques, contour detection, and morphological operators. The segmentation algorithms efficacy is evaluated with the support of a specialist, resulting in an 82% of efficiency in the segmentation of the lateral ventricles. On the other hand, the average processing time is 4.79 seconds. With these results, the study allows establishing the effectiveness of the automatic calculation of thresholds, as well as the use of contour extraction algorithms, and morphological operators as support tools that accentuates the validity of the segmentation model.