Abstract

A novel deep learning based model called Multi-Planar Spatial Convolutional Neural Network (MPS-CNN) is proposed for effective, automated segmentation of different sub-regions viz. peritumoral edema (ED), necrotic core (NCR), enhancing and non-enhancing tumor core (ET/NET), from multi-modal MR images of the brain. An encoder-decoder type CNN model is designed for pixel-wise segmentation of the tumor along three anatomical planes (axial, sagittal, and coronal) at the slice level. These are then combined, by incorporating a consensus fusion strategy with a fully connected Conditional Random Field (CRF) based post-refinement, to produce the final volumetric segmentation of the tumor and its constituent sub-regions. Concepts, such as spatial-pooling and unpooling are used to preserve the spatial locations of the edge pixels, for reducing segmentation error around the boundaries. A new aggregated loss function is also developed for effectively handling data imbalance. The MPS-CNN is trained and validated on the recent Multimodal Brain Tumor Segmentation Challenge (BraTS) 2018 dataset. The Dice scores obtained for the validation set for whole tumor (WT :NCR/NE +ET +ED), tumor core (TC:NCR/NET +ET), and enhancing tumor (ET) are 0.90216, 0.87247, and 0.82445. The proposed MPS-CNN is found to perform the best (based on leaderboard scores) for ET and TC segmentation tasks, in terms of both the quantitative measures (viz. Dice and Hausdorff). In case of the WT segmentation it also achieved the second highest accuracy, with a score which was only 1% less than that of the best performing method.

Highlights

  • Gliomas represent 40% of tumors of the Central Nervous System, and 80% of all malignant brain tumors

  • We have developed a deep learning based model called MultiPlanar Spatial Convolutional Neural Network (MPS-CNN), for the automated segmentation of brain tumors from multi-modal MR images

  • The encoder-decoder type ConvNet model for pixel-wise segmentation was found to perform better than other patch-based models, mainly due to the introduction of new concepts like spatial max-pooling and unpooling to preserve the spatial locations of the edge pixels while reducing segmentation error around the boundaries

Read more

Summary

INTRODUCTION

Gliomas (tumors of glial cells) represent 40% of tumors of the Central Nervous System, and 80% of all malignant brain tumors. The generative approach in references (Gooya et al, 2012) first computes the spatial a-priori or “atlas” from healthy brain MRI scans This is modified using an expectation maximization (EM) algorithm, over a given set of patient images, to detect the most likely localization of the tumor therein. An encoder-decoder type ConvNet model is designed for pixelwise segmentation of the tumor along three anatomical planes (axial, sagittal, and coronal) at the slice level These are combined, using a consensus fusion strategy with a fully connected Conditional Random Field (CRF) based postrefinement (Krähenbühl and Koltun, 2011), to produce the final volumetric segmentation of the tumor and its constituent subregions.

MATERIALS AND METHODS
Dataset
ConvNet for Tumor Segmentation
EXPERIMENTAL SETUP AND RESULTS
Experiment 1
Experiment 2
CONCLUSIONS
Findings
ETHICS STATEMENT
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.