Issues in using databases of pre-recorded physiological signals to test medical devices

Bench testing with databases of pre-recorded physiological signals is a common step in the verification and validation of medical devices. For example, in the case of an electrocardiographic (ECG) monitor to be tested, the physiological signals are ECG tracings that have previously been recorded from patients using other ECG monitors. Human overreaders then annotate the tracings to determine the occurrence of significant clinical events such as ventricular tachycardia. These annotations can be used to determine the sensitivity and specificity of the ECG monitor to be tested. This article aims to highlight some issues associated with such bench testing, such as the way the pre-recorded signals are presented to the medical device to be tested, the adequacy of single channel versus dual channel versus multi-channel testing in the bench testing, and the need for the human overreaders and the medical device to use the same set of rules for determining the occurrence of a clinical event. These issues are also applicable to other monitors such as the apnea monitor. Some suggestions for addressing the issues are presented.