Robust Automatic Rodent Brain Extraction Using Pulse-Coupled Neural Networks in 3D

Nigel Chou¹, Jolena Tan¹, Asad Abu Bakar Md Ali¹, Kai-Hsiang Chuang¹

We present an automatic brain-extraction algorithm optimized for rodents, based on a pulse-coupled neural network (PCNN) operating in 3D. PCNN links pixels with similar intensity, then a morphological operation is used to separate regions, of which the largest is selected as the brain mask. Using Jaccard index and True-positive Rate as a measures of similarity to a manual gold-standard, this method showed improved performance compared to an existing algorithm (Brain Surface Extraction) and a PCNN algorithm operating in 2D mode (on slices). Additional advantages include reduced user intervention and accurate segmentation of the olfactory bulb and paraflocculus of cerebellum.