Evaluation of Statistical Features for Medical Image Retrieval

In this paper we present a complete system allowing the classification of medical images in order to detect possible diseases present in them. The proposed method is developed in two distinct stages: calculation of descriptors and their classification. In the first stage we compute a vector of thirty-three statistical features: seven are related to statistics of the first level order, fifteen to that of second level where thirteen are calculated by means of co-occurrence matrices and two with absolute gradient; the last thirteen finally are calculated using run-length matrices. In the second phase, using the descriptors already calculated, there is the actual image classification. Naive Bayes, RBF, Support VectorMa- chine, K-Nearest Neighbor, Random Forest and Random Tree classifiers are used. The results obtained from the proposed system show that the analysis carried out both on textured and on medical images lead to have a high accuracy