Automated diagnosis of pneumoconiosis (black lung): A feasibility study

The feasibility of automatic diagnosis of pneumoconiosis using digital image analysis techniques was investigated. Normal/abnormal diagnosis was done first. Differential diagnosis was not attempted at this stage. Normal/abnormal classification scheme is based on a technique which makes use of individual lung lobe histograms. The magnitude of the first relative peak in the normalized histogram indicates the extent of the nodular patterns that characterize pneumoconiosis. For the abnormal lung lobe this value is higher than that for a normal one. This criterion is used for normal/abnormal classification. The results are encouraging enough to propose an automated pneumoconiosis diagnosis system. Differential diagnosis will be attempted by analyzing the nodular lung pattern. The size, number, and location of the nodularities will be used for differential diagnosis.