On the Implementation of Planar 3D Transfer Learning for End to End Unimodal MRI Unbalanced Data Segmentation

Abstract

This article describes detailed notes on the practical implementation of our paper Planar 3D transfer learning for end to end unimodal MRI unbalanced data segmentation (ICPR 2020, Milan), which deals with a problem of multiple sclerosis lesion segmentation from a unimodal MRI flair brain scan by applying a planar 3D transfer learning backbone weights to an autoencoder segmentation neural network. Our source code is published online under an open-source license, and we provide step-by-step instructions for the reproduction of our results.