Computer Aided Detection for Pneumoconiosis Screening on Digital Chest Radiographs

This paper presents a computer aided detection scheme on digital chest radiographs for pneumoconiosis screening. The scheme involves several medical image processing and analysis technologies, i.e. lung segmentation algorithm using the active shape model, image enhancement and features extraction from lung regions, feature down-selection by correlation analysis and clustering method, classification on lung regions by multi-scale classifiers support vector machine, and finally, chest based reporting out on the classification probabilities. Experiments are conducted on our digital chest xray database, and the result shows a good classification performance for screening application in clinic.