Computer aided diagnosis in thoracic and mammographic radiology
There has been a significant effort in the radiology department at Duke University to develop computer aided diagnosis (CAD) systems. The goal of the development of these systems is to assist radiologists in interpreting radiographic images and findings. These efforts have encompassed: (1) detection and size quantification of cold lesions in nuclear medicine (SPECT) images; (2) diagnosis of acute pulmonary embolism from physicians reading of ventilation/perfusion scans and chest radiographs; (3) prediction of breast cancer malignancy from mammographers' readings of diagnostic mammograms; (4) detection of pulmonary nodules from digital chest radiographs; (5) classification of interstitial lung disease; and (6) characterization of pulmonary nodules. Details of two of these efforts are described here: (1) prediction of breast cancer malignancy from mammographers' readings of diagnostic mammograms and (2) detection of pulmonary nodules from digital chest radiographs