[Paper Review] Brain Tumor Segmentation Using Deep Learning by Type Specific Sorting of Images

Zahra Sobhaninia, Safiyeh Rezaei+6 more|arXiv (Cornell University)|Sep 20, 2018

Brain Tumor Detection and Classification4 references72 citations

TL;DR

The paper investigates brain tumor segmentation using deep learning and compares a single-network approach to a multi-network, type-specific image sorting strategy, achieving higher Dice scores with multiple networks.

ABSTRACT

Recently deep learning has been playing a major role in the field of computer vision. One of its applications is the reduction of human judgment in the diagnosis of diseases. Especially, brain tumor diagnosis requires high accuracy, where minute errors in judgment may lead to disaster. For this reason, brain tumor segmentation is an important challenge for medical purposes. Currently several methods exist for tumor segmentation but they all lack high accuracy. Here we present a solution for brain tumor segmenting by using deep learning. In this work, we studied different angles of brain MR images and applied different networks for segmentation. The effect of using separate networks for segmentation of MR images is evaluated by comparing the results with a single network. Experimental evaluations of the networks show that Dice score of 0.73 is achieved for a single network and 0.79 in obtained for multiple networks.

Motivation & Objective

• Motivate accurate brain tumor segmentation due to critical diagnostic errors.
• Evaluate the impact of image type-specific sorting on segmentation performance.
• Compare a single-network approach with a multi-network approach for MR image segmentation.

Proposed method

• Apply deep learning to brain MR images for tumor segmentation.
• Experiment with different angles/views of MR images.
• Use separate networks for segmentation vs. a single unified network.
• Evaluate segmentation performance using Dice score as the primary metric.

Experimental results

Research questions

• RQ1 Does using separate networks for different image types improve segmentation accuracy over a single network?
• RQ2 What Dice score improvement is achieved by multi-network type-specific sorting compared to a single network?
• RQ3 Which image orientations/angles contribute most to segmentation performance?
• RQ4 Is there a trade-off between model complexity and segmentation accuracy in this setup.

Key findings

• Single-network Dice score: 0.73.
• Multiple networks with type-specific sorting: Dice score 0.79.
• Type-specific sorting and multi-network setup yields improved segmentation performance over a single network.