Towards Standardized Pen-Based Annotation of Breast Cancer Findings

Abstract

The development of computer technologies provides a means to support and facilitate the daily activities of potentially all users. This may be of particular importance for experts in breast cancer imaging and diagnosis. While many research efforts have been carried out separately on the implementation of task-oriented systems, much less effort has been undertaken to design and develop technologies compliant with domain standards or in accordance with end-user needs and expectations. This further suggests the need to improve both the usefulness and the usability of breast cancer-dedicated systems. This paper reports the results of a development method combining the application of usercentered design together with usability development methods. At different time frames in the life-cycle, the development method employed knowledge elicitation interviews, scenario-focused questionnaires, paper mock-ups and usability tests. Owing to its naturalness and its convenience, pen-based interaction with a graphics tablet was chosen as the modality to interact with the system. Additional innovative solutions were designed and implemented in order to facilitate and improve the visualization and the manipulation of data during the lesion characterization: namely an icon framework, a star-menu and a semi-automatic lesion detection system. The resulting user interface is a penbased interactive tool supporting visualization, navigation, standardized lesion characterization and reporting. The usability tests suggest that it provides endusers with an efficient, reliable and usable system. Context and MotivationsSince digital mammography has replaced screen-film mammography, information technology has progressively been introduced in breast cancer (BC) screening and diagnosis. Owing to the specificity and the complexity of each task involved in mammogram analysis (e.g., image visualization, image analysis, lesion detection, interpretation and reporting), research efforts have been focusing mainly on the implementation of task-oriented systems such as image viewers, computer-aided diagnosis (CAD) software, digital case databases, etc. Hence, radiologists currently