Multimedia files are increased in size and include affordable memory storage along with the popularity of World Wide Web (www) so that the requirement for effective tools for retrieving images from huge database is critical. For diagnosis, reference, therapy, surgery and training medical images are significant one. A meta-heuristic protocol is built for solving a vast variety of hard optimization issues with no need of deep adaptation to every issue. In this work, an automated Computed Tomography (CT) classification system with novel features selection technique on the basis of FireFly (FF) optimization protocol is suggested for medical images. Wavelets extract the features from medical images and selected CT image features are classified through usage of Naïve Bayes as well as k Nearest Neighbour. The suggested FF protocol attains improved accuracy in comparison to mutual information (MI) as well as the generic FF. Outcomes of simulations reveal that the suggested CFS-FF technique with NB enhanced classification accuracy by 2.83% than MI with NB and with kNN enhanced classification accuracy 2.86% than MI with kNN.