Stochastic analysis of ann statistical features for ct brain posterior fossa image classification

Abstract

Automated classification of Posterior Fossa (PF) in Computed Tomography (CT) image sequences is important for the diagnosing of ischemic. The identification of PF remains ambiguous due to its varying structures and regions. In this paper, a stochastic analysis of classification system is presented to classify PF slices in CT image sequences. This system can highlight the specified slices with credence, which can be a good visual companion piece for the radiologist to accurately identify the region with minimum processing time. The choice of eleven statistical features is based on the t-test and sensitivity analysis, which were conducted to investigate the significant features for Artificial Neural Network (ANN) predictive model. This classifier has been compared with the Support Vector Machine (SVM) and the effectiveness of both classifiers is leveraged through 10-fold cross-validation strategy. It is found that the ANN had achieved the highest classification accuracy of 95.1% in 1.695 s.