Lung tuberculosis identification based on statistical feature of thoracic X-ray

This paper presents experiments and results on lung tuberculosis (TB) identification by using computer. This research´s attempt is to reduce patient waiting time in obtaining X-ray diagnosis result on lung TB disease due to the mismatch the ratio of radiologist to the number of patients, especially in remote areas in Indonesia. To imitate radiologist which make visual examination on textural feature of thoracic X-ray images to make diagnosis, we exploit textural features calculated by computer to be used as descriptor in classifying images as TB or non-TB. We used statistical feature of image histograms by calculate five features: mean, standar deviation (std), skewness, kurtosis, and entropy. Features calculated where then reduced to two and one principal feature using Principal Componen Analysis (PCA) method. Finally, we used minimum distance classifier as classifier method based on two and one principal feature as descriptor. This experiment results shown that it is possible to classify TB and non-TB images based on statistical features on image histogram.