Abstract: The purpose of this dissertation is to diagnose solitary pulmonary nodules (SPN) as either benign or malignant both subjectively and objectively with features extracted from computed tomography (CT) images and a computer-aided diagnosis (CAD) system. There were a total of 81 patients with an SPN. These nodules were imaged with high resolution CT, 54 volumetrically, and 35 of those with contrast media. The diagnoses of all nodules were confirmed with pathology or radiographic followup. Each nodule was contoured semi-automatically forming a region of interest (ROI) twice and approved by a thoracic radiologist. One ROI contained the entire nodule including ground-glass components while the other contained only the solid portion. These two ROIs were subtracted to form a third ground-glass only ROI. Two- and three-dimensional features were extracted based on these ROIs. These features were designed to measure the size, shape, attenuation, and texture of the nodules. Subjectively, two observers studies were conducted, one at a conference with 81 readers; while another was conducted locally with 8 radiologists to assess the ability of readers to characterize nodules as benign or malignant. There were general trends showing observer performance improvement in this task with the addition of clinical information and output from a simple CAD system. To characterize SPNs objectively, the features were input into a number of different statistical classifiers. The output from these classifiers were then analyzed with receiver operating characteristic (ROC) techniques. The area under the ROC curves (A_Z) ranged from 0.68-0.92 when training and testing was done in a leave-one-out fashion and using two-dimensional ROIs. When using three-dimensional ROIs from nodules the A_Z ranged from 0.79-0.93. When using subjective input from the local observer study the range of A_Z was 0.75-0.95. The spatial distribution of iodinated contrast enhancement was investigated for 29 nodules by creating maps that indicate the spatial distribution of the enhancement pattern of nodules. These maps were investigated both qualitatively with three radiologists and objectively with texture analysis. Subjectively, the readers' A_Z ranged from 0.69-0.86 for the pattern of enhancement while the texture analysis yielded an A _Z of 0.84. SPNs were characterized as malignant or benign by a number of different approaches both qualitatively and quantitatively. The potential of CAD was demonstrated and its role in patient management by improving the diagnostic utility of CT imaging was alluded to.