Feature Extraction of Arterio-Venous Malformation Images using Grey Level Co-Occurrence Matrix

摘要

Objective: Arterio-Venous Malformation is, mostly, as a result of incidental findings. This is a pioneering effort to incorporate advanced image segmentation techniques in order to improve its diagnosis. Method: Feature extraction of Arterio-Venous Malformation (AVM) brain images is attempted for automatic AVM recognition system using OTSU based Particle Swarm Optimisation (PSO) algorithm. Initially the input Magnetic Resonance Image (MRI) is segmented using PSO method, based on multiple threshold processes. The image features are then extracted from the partitioned regions using Gray Level Co-occurrence Matrix (GLCM) technique. The features obtained from AVM and normal brain images are compared using statistical measures. Findings: Our analysis suggests that the extracted GLCM features of AVM brain MRI images shows significant variation to normal brain MRI images. Out of the total 22 features extracted, 18 features shows a lesser feature extraction value for AVM affected brain image compared to the normal brain image. The other 4 features extracted shows a higher value for AVM affected brain image compared to the normal brain image. This pattern is the same for any AVM affected brain, in comparison to a normal brain. Applications: This work helps in the development of automatic recognition system for AVM, so that many cases can be identified in the preliminary stages and suitably treated.