Brain extraction on MRI scans in presence of diffuse glioma: Multi-institutional performance evaluation of deep learning methods and robust modality-agnostic training

Siddhesh Thakur,Gaurav Shukla,Pamela Lamontagne,Sarthak Pati,Arash Nazeri,Gregory S Alexander,Christos Davatzikos,Saima Rathore,Spyridon Bakas,Mikhail Milchenko,Joseph Lombardo,Jimit Doshi,Aikaterini Kotrotsou,Joshua D Palmer,Sung Min Ha,Michel Bilello,Chiharu Sako,Spencer Liem,Uday Kulkarni,Sanjay N Talbar ,Rivka R Colen ,Adam E Flanders ,Güray Erus ,Daniel S Marcus

doi:10.1016/j.neuroimage.2020.117081

Abstract

Brain extraction, or skull-stripping, is an essential pre-processing step in neuro-imaging that has a direct impact on the quality of all subsequent processing and analyses steps. It is also a key requirement in multi-institutional collaborations to comply with privacy-preserving regulations. Existing automated methods, including Deep Learning (DL) based methods that have obtained state-of-the-art results in recent years, have primarily targeted brain extraction without considering pathologically-affected brains. Accordingly, they perform sub-optimally when applied on magnetic resonance imaging (MRI) brain scans with apparent pathologies such as brain tumors. Furthermore, existing methods focus on using only T1-weighted MRI scans, even though multi-parametric MRI (mpMRI) scans are routinely acquired for patients with suspected brain tumors. In this study, we present a comprehensive performance evaluation of recent deep learning architectures for brain extraction, training models on mpMRI scans of pathologically-affected brains, with a particular focus on seeking a practically-applicable, low computational footprint approach, generalizable across multiple institutions, further facilitating collaborations. We identified a large retrospective multi-institutional dataset of n=3340 mpMRI brain tumor scans, with manually-inspected and approved gold-standard segmentations, acquired during standard clinical practice under varying acquisition protocols, both from private institutional data and public (TCIA) collections. To facilitate optimal utilization of rich mpMRI data, we further introduce and evaluate a novel ‘‘modality-agnostic training’’ technique that can be applied using any available modality, without need for model retraining. Our results indicate that the modality-agnostic approach11Publicly available source code: https://github.com/CBICA/BrainMaGe obtains accurate results, providing a generic and practical tool for brain extraction on scans with brain tumors.

Highlights

Brain extraction, known as skull-stripping, describes the process of removing the skull and non-brain tissues from brain magnetic resonance imaging (MRI) scans
We note that the best performing Deep Learning (DL) model for brain extraction in T1 brain tumor scans (2D-Res-Inc - Fig. 5) ends up with the lowest performance among the DL architectures when trained on healthy brain T1 scans (2D-Res-Inc-H - Fig. 5), with its Dice Similarity Coefficient (Dice) showing a decrease of at least 6% across the data from the multiple institutions
Brain extraction is a key component of computer algorithms that are poised to change the practice of clinical neuroradiology

Summary

Introduction

Known as skull-stripping, describes the process of removing the skull and non-brain tissues (e.g., neck fat) from brain magnetic resonance imaging (MRI) scans. It is a crucial step for preprocessing neuro-imaging datasets and has an immediate bearing on all subsequent investigative procedures. It is a necessary processing step in most studies for compliance with privacy-preserving regulations, such as the Health Insurance Portability and Accountability Act of 1996 (HIPPA) and the General Data Protection Regulation of 2016 (GDPR). DL methods, and Convolutional Neural Networks (CNNs), have obtained state-of-the-art results in multiple problems of image segmentation

Methods

Results

Discussion

Conclusion