Abstract: Automated tuberculosis detection systems will help radiologists to diagnose the disease easily. The proposed system implements an advanced detection system using chest radiographs. The system uses a minimum cross entropy segmentation for extracting the lung regions from chest radiographs. After the segmentation process several features are extracted for the classification stage. A probabilistic neural network was used for the classification. The images are classified as normal or abnormal by the classifier.
Keywords: Tuberculosis, minimum cross entropy, computer aided diagnosis, chest X-ray, tamura.