Abstract

Embedded micro-processors are becoming increasingly important in medical treatments. These devices are embedded in CAT scanners, radiation therapy devices, MRIs, and are used in the control of drug production. Embedded systems are becoming more sophisticated, and the programs that control these system have increasing demands for precision and reliability. Recently, the FDA hasbeen under increased pressure by Congress to end the paralysis within the agency in granting pre-market approval to the so-called "high-risk" devices. Complaints have been aimed at the FDA's Center for Devices and Radiological Health, which has become reluctant to approve devices. This fear, it is claimed, is due to several past controversies including the silicone-breast implants and the Pfizer Inc. heart valve. As evidence that there really is a slow-down in approval, in fiscal 1986, 72 approvals were granted, in fiscal 1991, 27 approvals were granted, and in scal 1992, only 12 premarket high-risk approvals were granted [1]. In this paper, we will briefly present a software engineering technique called sensitivity analysis that aims at producing software that is less likely to hide faults. When sensitivity analysis indicates that faults are not likely to hide from testing, we can more comfortably rely on the results of testing. Sensitivity analysis has been automated, and preliminary results from the automated analysis suggest that the technique will have significant advantages for the development, validation, and regulation of medical devices.

Citations