An Interactive Task Analysis Framework and Interactive System Research for Computer Aided Diagnosis

Computer aided detection and diagnosis system (CAD) is designed to improve the efficiency and quality of medical diagnosis. The existing CAD system tries to realize a diagnosis process in an automatic way to replace the doctor's diagnosis work completely, but it is difficult to capture the doctor's detection and diagnosis needs and goals accurately due to lack of doctor's experience. Thus, the accuracy is not up to the doctor's manual diagnosis results, and it is a tedious and cumbersome job to use the system because it requires manual intervention, which leads to the fact that CAD system was difficult to be widely used. This paper analyzes the characteristics of doctors' diagnosis tasks deeply, and makes the reasonable allocation of tasks between the doctor and computer. Then the perceptual control theory is introduced into the task model, a PTT task model is proposed to make it express uncertainty of the doctor's diagnosis tasks, and it can be used as a reference for future uncertainty task analysis. Therefore, an interactive CAD tasks framework is put forward with the assistant decision support provided for doctors by means of Bayesian network model and conceptual graph semantic matching method to diagnose tasks. In addition, effective support is provided for interface development staffs to design interactive systems that are suitable for the physicians' diagnostic tasks. The experimental results show that the method not only improves the accuracy and efficiency of the medical diagnosis, but also improves the usability of the system design.