The effect of varying high pass filter parameters on the detectability of isolated pulmonary nodules in digitized chest images

The effect of image processing of digital radiographic data on lesion detectability in clinical images has not been systematically studied. In this experiment, we evaluated a low frequency suppression filter program applied to CT digital radiographic localization images (General Electric Scoutview). Three different filter parameters affecting edge enhancement and local image contrast were applied to a set of digital chest images. A standard observer detection experiment comparing the variously filtered digital images and standard chest radiographs in the detection of lung nodules (11 peripheral lung fields, one superimposed on aortic arch) was performed. Standard chest radiography was more sensitive than the digital chest images, although some improvement was noted with increase in local image contrast and edge enhancement. Both image formats were equally specific. Image processing would be better evaluated using a digital imaging system with better performance parameters.