The usage of algorythm based on color descriptor in the laparoscopic images classification

The classification of laparoscopic images in correspondence to certain pathological state of the abdominal organs assumes the exploration of characteristic of color descriptor of abdominal structures which are suspected for the pathology presence. Sparse publications which were devoted to computer automatic diagnostics (CAD) in the laparoscopic surgery till the recent time have not been dealt with the color descriptors as an informative one for diagnostics. That is why the aim of the work is to investigate the effectiveness of the usage of algorithm of the objects classification, when cascade classifier is learned with color descriptors. The general scheme of CAD was based on the realization of algorithm which supposed to capture video frame with the next gamma-correction and conversion of the color scale from RGB to HSV, extraction and selection of descriptors with the final classification of images. The uttermost was performed with the usage of database of proper images and technologies of learning in accordance to cascade classifier learning principles. The gained data revealed that created CAD was able to gain most pronounced true positive diagnoses in cases of liver cysts and liver metastasis which were registered in 84,0 % and 68,0 % of all cases correspondently.