| The chances of surviving a breast cancer diagnosis as well as the effectiveness of any potential treatments increase significantly with early detection of the disease. As such, a considerable amount of research is being conducted to augment the breast cancer detection and diagnosis process. One such area of research involves the investigation and application of sophisticated computer algorithms to assist clinicians in detecting and diagnosing breast cancer on medical images (termed generally as "computer-aided diagnosis" or CAD).;This study investigated a previously-developed breast ultrasound CAD system with the intent of translating it into a clinically-viable system. While past studies have demonstrated that breast ultrasound CAD may be a beneficial aid during the diagnosis of breast cancer on ultrasound, there are no investigations concerning its potential clinical translation and there are currently no commercially-available implementations of such systems. This study "bridges the gap" between the laboratory-developed system and the steps necessary for clinical implementation.;A novel observer study was conducted that mimicked the clinical use of the breast ultrasound CAD system in order to assess the impact it had on the diagnostic performance of the user. Several robustness studies were also performed: the sonographic features used by the system were evaluated and the databases used for calibration and testing were characterized, the effect of the user's input was assessed by evaluating the performance of the system with variations in lesion identification and image selection, and the performance of the system on different patient populations was investigated by evaluating its performance on a database consisting solely of patients with Asian ethnicity.;The analyses performed here indicate that the breast ultrasound CAD system under investigation is robust and demonstrates only minor variability when subjected to "real-world" use. All of these results are critical to the success of a CAD system in a clinical environment, and they support the execution of a prospective clinical trial of the system. |
