Segmentation of functional MRI by K-means clustering

A preliminary study was conducted to segment 1.5T fMRIs into the microvasculature and relatively large blood vessels using the intensity, phase and temporal delay of activated pixels as three correlated parameters in gradient echo images. Images acquired during visual stimulation using a checkerboard flashing at 8Hz were investigated. Activated pixels were identified by correlating their time-course in the gradient echo images to a model of the stimulation 'on'-'off' sequence. The temporal delay of each activated pixel was estimated by fitting its time-course to a reference sinusoidal function. The mean signal intensity and phase difference of the activated pixels was computed by subtracting the average of the 'on' images from the average of the 'off' images. After mapping each pixel onto a three-dimensional feature space (intensity, phase shift and temporal delay), a clustering method based on a K-means algorithm was employed to classify pixels into two or three classes representing the relatively large blood vessels and the microvasculature. Good demarcation between large veins and activated gray matter was achieved with this method.