CLASSIFICATION OF BRAIN HEMORRHAGE (CT SCANS) USING TEXTURAL FEATURE AND CNN DEEP LEARNING

This study acts in image processing and analysis in computerized-aid so as to increase the accuracy of diagnosis in reporting the classification of the brain haemorrhage type. To do image processing, one needs many techniques but there is a need to offer an improvement for the algorithm of image processing to expand preciseness. This project provides an algorithm for CT scan image. The way has three processes: Segmentation (parsing into small parts), feature extraction (knowing features), classification (reporting) using Neural Network of MATLAB. Reconstruction was used before to do the current project process but it was time-consuming. Reconstructed images change their features many times so it is difficult to use automation here because of different types of brain bleeding and their shapes. Textural features are entropy (random), mean, variance, inverse. They are calculated on gray-level co-occurrence matrix. The algorithm given is used to classify one of four brain bleedings: Cerebellar (back brain part), Extra-dural(outside brain case), Intraventricular (brain cavity area), Subdural (between brain and skull). The experiment conducted in Elrazi University in Sudan Khartoum city, besides the CT scans images collected from AlSaahah Hospital in Sudan in Khartoum city and others images from the internet.