Outlier detection in the design of fault-tolerant biosensor arrays

Abstract

The possibility of fabricating a digital voter that will detect and eliminate a faulty sensor in an array of identical biosensors is examined. Eleven statistical outlier-detection procedures are applied to the responses of an array of antimony-antimony oxide penicillinase electrodes and to an extensive computer simulation of small array responses. A Dixon excess-over-range test and a maximum normalized residual test are shown to be safe outlier-detection procedures that will detect a faulty sensor and offer an algorithm that may be economically implemented in hardware. The Iglewicz & Martinez test, which can be implemented more conveniently in software, is shown to be very efficient when applied to the real data. However, its poorer performance when applied to the simulated data suggests that further examination of this test is required.