Knowledge-driven automated detection of pleural plaques and thickening in high resolution CT of the lung.

Consistent efforts are being made to build Computer-Aided Detection and Diagnosis systems for radiological images. Such systems depend on automated detection of various disease patterns, which are then combined together to obtain differential diagnosis. For diffuse lung diseases, over 12 disease patterns are of interest in High Resolution Computed Tomography (HRCT) scans of the lung. In this paper, we present an automated detection method for two such patterns, namely Pleural Plaque and Diffuse Pleural Thickening. These are characteristic features of asbestos-related benign pleural disease. The attributes used for detection are derived from anatomical knowledge and the heuristics normally used by radiologists, and are computed automatically for each scan. A probabilistic model built on the attributes using naïve Bayes classifier is applied to recognise the features in new scans, and preliminary results are presented. The technique is tested on 140 images from 13 studies and validated by an experienced radiologist.