A CAD System for the Early Prediction of Hypertension based on Changes in Cerebral Vasculature

Document Type

Conference Proceeding

Source of Publication

IST 2019 - IEEE International Conference on Imaging Systems and Techniques, Proceedings

Publication Date

12-1-2019

Abstract

© 2019 IEEE. Hypertension is a leading cause for mortality in the US and a significant contributor to many vascular and non vascular diseases. Previous literature reports suggest that specific cerebral vascular alterations precede the onset of hypertension. In this manuscript, we propose a magnetic resonance angiography (MRA)-based computer-aided-diagnosis (CAD) system for the early detection of hypertension. The steps of the proposed CAD system are: 1) preprocessing of the MRA input data to correct the bias resulting from the magnetic field, remove noise effects, reduce contrast non-uniformities, enhance homogeneity using a generalized Gauss-Markov random field (GGMRF), and normalize data to enhance the segmentation process, 2) delineating the cerebral vasculature using a deep 3-D convolutional neural network (CNN) automatically and accurately, 3) extraction of vascular features (cerebrovascular diameters and tortuosity) that are reported to change with the progression of hypertension and constructing the feature vectors, 4) using the feature vectors for classifying input data using a support vector machine (SVM) classifier. We report a 90% classification accuracy in distinguishing between normal and potential hypertensive subjects. These results demonstrate the efficacy of using the proposed vascular features to predict pre-hypertension or hypertension. Clinicians could track the alterations of these vascular features over time for people at risk of developing hypertension for optimal medical management and mitigate adverse events.

ISBN

9781728138688

Publisher

Institute of Electrical and Electronics Engineers Inc.

Last Page

5

Disciplines

Medicine and Health Sciences

Keywords

Blood Vessels, Cerebral, CNN, Hypertension, SVM, Tortuosity

Scopus ID

85081987701

Indexed in Scopus

yes

Open Access

no

Share

COinS