Accurate Segmentation of Cerebrovasculature From TOF-MRA Images Using Appearance Descriptors

Abstract

Analyzing cerebrovascular changes can significantly lead to not only detecting the presence of serious diseases e.g., hypertension and dementia, but also tracking their progress. Such analysis could be better performed using Time-of-Flight Magnetic Resonance Angiography (ToF-MRA) images, but this requires accurate segmentation of the cerebral vasculature from the surroundings. To achieve this goal, we propose a fully automated cerebral vasculature segmentation approach based on extracting both prior and current appearance features that have the ability to capture the appearance of macro and micro-vessels in ToF-MRA. The appearance prior is modeled with a novel translation and rotation invariant Markov-Gibbs Random Field (MGRF) of voxel intensities with pairwise interaction analytically identified from a set of training data sets. The appearance of the cerebral vasculature is also represented with a marginal probability distribution of voxel intensities by using a Linear Combination of Discrete Gaussians (LCDG) that its parameters are estimated by using a modified Expectation-Maximization (EM) algorithm. The extracted appearance features are separable and can be classified by any classifier, as demonstrated by our segmentation results. To validate the accuracy of our algorithm, we tested the proposed approach on invivo data using 270 data sets, which were qualitatively validated by a neuroradiology expert. The results were quantitatively validated using the three commonly used metrics for segmentation evaluation: the Dice coefficient, the modified Hausdorff distance, and the absolute volume difference. The proposed approach showed a higher accuracy compared to two of the existing segmentation approaches.

BRAVE-NET: Fully Automated Arterial Brain Vessel Segmentation in Patients With Cerebrovascular Disease

“…First, the majority of MRA-based vessel segmentation frameworks presented so far in the literature require image pre-processing, such as downsampling, brain masking, intensity correction, image normalization, and various other methods before they can be applied (Lesage et al, 2009 ; Chen et al, 2017 ; Livne et al, 2019a ; Ni et al, 2020 ; Taher et al, 2020 ). Applications in computed tomography angiography (CTA) do not differ in that regard and include for example deformable matching, atlas co-registration, candidate vessel selection and feature extraction (Passat et al, 2005 ; Meijs et al, 2017 ).…”

“…As a comparison, predictions for the same images on a standard GPU (NVIDIA Titan Xp) system take 40 s. Second, previous frameworks were developed using smaller datasets ranging from 10 to 100 patient scans derived from a single scanner (Passat et al, 2005 ; Wang et al, 2015b ; Chen et al, 2017 ; Livne et al, 2019a ; Ni et al, 2020 ; Patel et al, 2020 ). Taher et al ( 2020 ) used a patient cohort size ( N = 270) similar to our study.…”

“…Consequently, such models usually perform worse on datasets differing from the dataset distribution used for training. For example, Taher et al ( 2020 ) used the freely available model of Livne et al ( 2019a , b ) for vessel segmentation on their own data and noticed a reduced performance compared to the original publication. Here, the increased generalizability of the model presented in this work is supported by its ability to capture a considerably greater degree of data variation caused both by larger technical and pathological variance in the datasets.…”

BRAVE-NET: Fully Automated Arterial Brain Vessel Segmentation In Patients with Cerebrovascular Disease

“…Second, previous frameworks were developed using smaller datasets ranging from 10 to 100 patient scans derived from a single scanner (Passat et al, 2005;Wang et al, 2015b;Chen et al, 2017, 3;Livne et al, 2019a;Ni et al, 2020;Patel et al, 2020). Taher et al used a patient cohort size (N=270) similar to our study (Taher et al, 2020). Third, the majority of reported models are based on data obtained from healthy individuals and did not include patients with (vessel) pathologies.…”

“…First, the majority of MRA-based vessel segmentation frameworks presented so far in the literature require image pre-processing such as downsampling, brain masking, intensity correction, image normalization, and various other methods before they can be applied (Lesage et al, 2009; Chen et al, 2017; Livne et al, 2019a; Ni et al, 2020; Taher et al, 2020). Applications in computed tomography angiography (CTA) do not differ in that regard and include for example deformable matching, atlas co-registration, candidate vessel selection and feature extraction (Passat et al, 2005; Meijs et al, 2017).…”