We present two well-known statistical methods here to extract features related to different lung diseases on the chest X-ray images: 1).Texture feature; 2). Moment parameter feature. 11 parameters of texture feature and 7 moment parameters among many others were chosen to represent the characteristics of three different lung diseases which are frequently encountered in routine clinical diagnoses and as features for further recognition use. At the recognition step, the K-means algorithm based on the least-square function was employed as a classifier to distinguish 18 samples based on those parameters. Recognition rates from about 70% to 100% are achieved according to different diseases.
F. Zhou, F. Zhou,
"Computer aided diagnoses of lung diseases through radiographs", Proc. SPIE 1395, Close-Range Photogrammetry Meets Machine Vision, 13954C (1 August 1990); doi: 10.1117/12.2294399; https://doi.org/10.1117/12.2294399