Digitization of films and texture analysis for digital classification of pulmonary opacities
Desaga, J.F.; Dengler, J.; Wolf, T.; Engelmann, U.; Scheppelmann, D.; Meinzer, H.P.
Roentgen-Blaetter 41(4): 147-151
The study aimed at evaluating the effect of different methods of digitisation of radiographic films on the digital classification of pulmonary opacities. Test sets from the standard of the international Labour Office (ILO) Classification of Radiographs of Pneumoconiosis were prepared by film digitisation using a scanning microdensitometer or a video digitiser based on a personal computer equipped with a real time digitiser board and a vidicon or a Charge Coupled Device (CCD) camera. Seven different algorithms were used for texture analysis resulting in 16 texture parameters for each region. All methods used for texture analysis were independent of the mean grey value level and the size of the image analysed. Classification was performed by discriminant analysis using the classes from the ILO classification. A hit ratio of at least 85% was achieved for a digitisation by scanner digitisation or the vidicon, while the corresponding results of the CCD camera were significantly less good. Classification by texture analysis of opacities of chest X-rays of pneumoconiosis digitised by a personal computer based video digitiser and a vidicon are of equal quality compared to digitisation by a scanning microdensitometer. Correct classification of 90% was achieved via the described statistical approach.