Detection of Polymicrogyria

This study introduces the open pediatric MRI dataset (PPMR), which includes both polymicrogyria (PMG) and control cases from the Children’s Hospital of Eastern Ontario. The subtle differences between PMG and control MRIs make automatic detection challenging, but the study proposes an anomaly detection method using a novel center-based deep contrastive metric learning loss function (cDCM).The method achieved an 88.07% recall at 71.86% precision, representing a pioneering application of machine learning for identifying PMG solely from MRI scans.