Abstract #1006

Graham J, Ang S, Allen P, Hutchinson C

This study investiagated the use of image texture analysis in classification of Benign Prostatic Hyperplasia as stromal or non-stromal. The distinction has important implication for drug treatment and can only be reliably made by histology. Texture measurements based on grey-level co-occurrence matrices were taken from the central gland regions of T2-weighted fat-suppressed images from 19 patients with BPH. Several image features produced classification into the two groups that agreed significantly (values in the range 0.14 to 0.41) with a semi-quantitative visual assessment based on published criteria that has been shown to accurately correspond to histological type.