Diagnostically Useful Video Content Extraction for Integrated Computer-Aided Bronchoscopy Examination System

The problem of diagnostically important content selection in bronchoscopy video was the subject of our research reported in this paper. The characteristic of illegible and redundant bronchoscopy video content was presented and analyzed. A method for diagnostically important frame extraction from non-informative, diagnostically useless content was proposed. Our methodology exploits region of interests segmentation, features extraction in multiresolution wavelet domain and frame classification. SVM with optimized kernels and quality criteria was applied for classification. Effectiveness of proposed method was verified experimentally on large image dataset containing about 1500 diversified video frames from different bronchoscopy examinations. Obtained results with mean sensitivity above 97% and mean specificity about 94% confirmed high effectiveness of proposed method.