Blue Room - Session 4
2:20 to 3:20 p.m. Tuesday, June 3, 2014
Big Data and Medical Device Reliability
When software is embedded into medical devices, how do we know that the software will be reliable? Software reliability is typically assessed by fitting various models to software defect data; the defects are found by executing software or system reliability tests. Reliability assessments made through this process are accurate only to the degree that the software reliability testing represents how the software is actually used. In many cases, we do not know exactly how software will be used, and therefore we cannot accurately predict its reliability. The concept of an operational profile has been introduced and applied to address this issue. However, operational profiles are limited in their ability to demonstrate extremely high reliability levels, which are typically required for medical devices. The explosion of device and patient data collected within medical devices, which has occurred in recent years, has created a potential to use this data in reliability testing. This presentation will: 1) Explain the weaknesses in current software reliability testing. 2) Propose a new approach to leverage big data in reliability testing of medical device software. 3) Describe a system that results in a highly resolute and model-free reliability assessment. 4) Introduce a new role of the software reliability engineer leading this effort.
Key Words: Software Reliability, Medical Device Reliability, Medical Device Safety, Model-free Reliability, Empirical Reliability, Reliability Assessment, Safety Assessment
Gary Berg and Chao Wang