Multi-Disease Segmentation of Gliomas and White Matter Hyperintensities in the BraTS Data Using a 3D Convolutional Neural Network

Jeffrey D Rudie,Rachit Saluja,Andreas M Rauschecker,John B Colby,Jiancong Wang,Leo Sugrue,Spyridon Bakas,David A Weiss

doi:10.3389/fncom.2019.00084

Abstract

An important challenge in segmenting real-world biomedical imaging data is the presence of multiple disease processes within individual subjects. Most adults above age 60 exhibit a variable degree of small vessel ischemic disease, as well as chronic infarcts, which will manifest as white matter hyperintensities (WMH) on brain MRIs. Subjects diagnosed with gliomas will also typically exhibit some degree of abnormal T2 signal due to WMH, rather than just due to tumor. We sought to develop a fully automated algorithm to distinguish and quantify these distinct disease processes within individual subjects’ brain MRIs. To address this multi-disease problem, we trained a 3D U-Net to distinguish between abnormal signal arising from tumors vs. WMH in the 3D multi-parametric MRI (mpMRI, i.e., native T1-weighted, T1-post-contrast, T2, T2-FLAIR) scans of the International Brain Tumor Segmentation (BraTS) 2018 dataset (ntraining = 285, nvalidation = 66). Our trained neuroradiologist manually annotated WMH on the BraTS training subjects, finding that 69% of subjects had WMH. Our 3D U-Net model had a 4-channel 3D input patch (80 × 80 × 80) from mpMRI, four encoding and decoding layers, and an output of either four [background, active tumor (AT), necrotic core (NCR), peritumoral edematous/infiltrated tissue (ED)] or five classes (adding WMH as the fifth class). For both the four- and five-class output models, the median Dice for whole tumor (WT) extent (i.e., union of AT, ED, NCR) was 0.92 in both training and validation sets. Notably, the five-class model achieved significantly (p = 0.002) lower/better Hausdorff distances for WT extent in the training subjects. There was strong positive correlation between manually segmented and predicted volumes for WT (r = 0.96) and WMH (r = 0.89). Larger lesion volumes were positively correlated with higher/better Dice scores for WT (r = 0.33), WMH (r = 0.34), and across all lesions (r = 0.89) on a log(10) transformed scale. While the median Dice for WMH was 0.42 across training subjects with WMH, the median Dice was 0.62 for those with at least 5 cm3 of WMH. We anticipate the development of computational algorithms that are able to model multiple diseases within a single subject will be a critical step toward translating and integrating artificial intelligence systems into the heterogeneous real-world clinical workflow.

Highlights

A significant challenge in the deployment of advanced computational methods into typical clinical workflows is the vast heterogeneity of disease processes, which are present both between individuals and within individuals
We used a 3D convolutional neural networks (CNNs) (U-Net architecture; Cicek et al, 2016; Milletari et al, 2016) for multiclass tissue segmentation with performance at the top 10% of the Brain Tumor Segmentation (BraTS) 2018 leaderboard (Bakas et al, 2019; noting that we did not participate in the official competition)
U-Nets have been adept at medical image segmentation, due to their ability to convert feature maps obtained during convolutions into a vector and from that vector reconstruct a segmentation, which reduces distortion by preserving the structural integrity

Summary

Introduction

A significant challenge in the deployment of advanced computational methods into typical clinical workflows is the vast heterogeneity of disease processes, which are present both between individuals (inter-subject heterogeneity) and within individuals (intra-subject heterogeneity). Most adults over the age of 60 have a variable degree of abnormal signal on brain MRIs due to age-related changes manifesting as white matter hyperintensities (WMH), which are typically secondary to small vessel ischemic disease (SVID) and chronic infarcts that can be found in subjects with vascular risk factors and clinical histories of stroke and dementia (Wardlaw et al, 2015) These lesions can confound automated detection and segmentation of other disease processes, including brain tumors, which result in abnormal signal in T2-weighted (T2) and T2 Fluidattenuated inversion recovery (T2-FLAIR) MRI scans secondary to neoplastic processes and associated edema/inflammation. Deep learning (DL) approaches for biomedical image segmentation are established as superior to the previous generation of atlas-based and hand-engineered feature approaches (Fletcher-Heath et al, 2001; Gooya et al, 2012), as demonstrated by their performance in recent image segmentation challenges (Chang, 2017; Kamnitsas et al, 2017; Li et al, 2018; Bakas et al, 2019; Myronenko, 2019)

Methods

Results

Conclusion