An ROC study detecting ability of idiopathic pulmonary fibrosis using digital radiography

One potential advantage of the digital radiography system is its ability to enhance image quality by various types of processing. Digital unsharp masking is one of the simplest and most useful forms of enhancing processes. The efficacy of unsharp masking in radiological diagnosis has not been investigated thoroughly. To evaluate the effects of unsharp masking in film-digital chest images, 3 observers were shown 150 test radiographs. These test radiographs consisted of 50 unprocessed images (25 normals and 25 patients with idiopathic pulmonary fibrosis with honey combing) and their 100 processed images by using 450 and 15-sized masks respectively. An ROC analysis of these data suggests that unsharp masking is more effective in detecting idiopathic pulmonary fibrosis than unprocessed image (ρ < 0.05), and so it may improve diagnostic accuracy for interstitial fibrosis. In addition, the smaller mask size (15) is more effective than the larger one (mask size 45) (ρ < 0.05). By using this analytic approach, an optimal parameter in digital chest radiography may be investigated in many other forms of pulmonary disease such as pulmonary nodule or mediastinal mass