DeepVesselNet: Vessel Segmentation, Centerline Prediction, and Bifurcation Detection in 3-D Angiographic Volumes

Abstract

We present DeepVesselNet, an architecture tailored to the challenges faced when extracting vessel trees and networks and corresponding features in 3-D angiographic volumes using deep learning. We discuss the problems of low execution speed and high memory requirements associated with full 3-D networks, high-class imbalance arising from the low percentage (<3%) of vessel voxels, and unavailability of accurately annotated 3-D training data—and offer solutions as the building blocks of DeepVesselNet. First, we formulate 2-D orthogonal cross-hair filters which make use of 3-D context information at a reduced computational burden. Second, we introduce a class balancing cross-entropy loss function with false-positive rate correction to handle the high-class imbalance and high false positive rate problems associated with existing loss functions. Finally, we generate a synthetic dataset using a computational angiogenesis model capable of simulating vascular tree growth under physiological constraints on local network structure and topology and use these data for transfer learning. We demonstrate the performance on a range of angiographic volumes at different spatial scales including clinical MRA data of the human brain, as well as CTA microscopy scans of the rat brain. Our results show that cross-hair filters achieve over 23% improvement in speed, lower memory footprint, lower network complexity which prevents overfitting and comparable accuracy that does not differ from full 3-D filters. Our class balancing metric is crucial for training the network, and transfer learning with synthetic data is an efficient, robust, and very generalizable approach leading to a network that excels in a variety of angiography segmentation tasks. We observe that sub-sampling and max pooling layers may lead to a drop in performance in tasks that involve voxel-sized structures. To this end, the DeepVesselNet architecture does not use any form of sub-sampling layer and works well for vessel segmentation, centerline prediction, and bifurcation detection. We make our synthetic training data publicly available, fostering future research, and serving as one of the first public datasets for brain vessel tree segmentation and analysis.

Automated Quantitative Analysis of Blood Flow in Extracranial–Intracranial Arterial Bypass Based on Indocyanine Green Angiography

“…Recently, deep learning methods have been widely used in medical image segmentation (36), among which UNet is the most commonly used for vessel segmentation (15). UNet has been extensively used in retinal vessel segmentation (37)(38)(39)(40), 3D cerebrovascular segmentation (41)(42)(43), and cardiac vessel segmentation. Sevastopolsky et al (44) applied UNet to segment the optic disc and cup in retinal fundus images to diagnose glaucoma.…”

A Novel and Open-Source Application for Vasculature Dataset Analysis and Visualization

“…To test our filters, a set of synthetically generated vasculature datasets with branchpoint labels were analyzed and results were compared to previous programs 16,28 . Labeled branchpoints from these datasets were considered as ground truths.…”

“…Several of these tools can extract 3D network characteristics, yet they depend on simple centerline analyses, provide limited feature output, and markedly over-label branchpoints, producing inaccurate results 7,16 . Other modern analysis packages that extract more accurate 6,17 and detailed features 18,19 from 3D vascular networks require experience with programming or interaction with terminals, potentially leading to unwelcome usage barriers and steep learning curves for researchers. Many publications also make use of private code or proprietary software for feature extractions [20][21][22] .…”

Automated Segmentation of the Human Abdominal Vascular System Using a Hybrid Approach Combining Expert System and Supervised Deep Learning

“…There is a growing interest in applications of artificial intelligence in cardiovascular diseases [ 4 , 9 , 10 , 11 ]. Emerging techniques, including machine learning (ML), and deep learning (DL), especially convolutional neural networks (CNN), have brought new insights into cardiovascular image segmentation [ 4 , 9 , 11 , 12 , 13 , 14 , 15 ] and could be applied to develop a fully automatic software. Synthetic data refer to data that are generated by a computer program, instead of being extracted from direct measurement by a human.…”