Support vector machine classification of arterial volume-weighted arterial spin tagging images.

ABSTRACT: INTRODUCTION:In recent years, machine-learning techniques have gained growing popularity in medical image analysis. Temporal brain-state classification is one of the major applications of machine-learning techniques in functional magnetic resonance imaging (fMRI) brain data. This article explores the use of support vector machine (SVM) classification technique with motor-visual activation paradigm to perform brain-state classification into activation and rest with an emphasis on different acquisition techniques. METHODS:Images were acquired using a recently developed variant of traditional pseudocontinuous arterial spin labeling technique called arterial volume-weighted arterial spin tagging (AVAST). The classification scheme is also performed on images acquired using blood oxygenation-level dependent (BOLD) and traditional perfusion-weighted arterial spin labeling (ASL) techniques for comparison. RESULTS:The AVAST technique outperforms traditional pseudocontinuous ASL, achieving classification accuracy comparable to that of BOLD contrast images. CONCLUSION:This study demonstrates that AVAST has superior signal-to-noise ratio and improved temporal resolution as compared with traditional perfusion-weighted ASL and reduced sensitivity to scanner drift as compared with BOLD. Owing to these characteristics, AVAST lends itself as an ideal choice for dynamic fMRI and real-time neurofeedback experiments with sustained activation periods.

SUBMITTER: Shah YS 

PROVIDER: S-EPMC5167003 | biostudies-literature | 2016 Dec

REPOSITORIES: biostudies-literature