Comments on "Optimal multiple level decision fusion with distributed sensors"

In the original article ~see ibid vol. 29, no. 4, p. 1252-59, 1993\ the authors consider the problem of target detection with multiple sensors and propose a decentralized decision network with multiple fusion levels with improved reliability to link failures. They also derive the detection performance of the system, called multiple level parallel decision fusion (MLPDF), and carry out a comparison of their system with the conventional parallel distributed model. In this letter we show that, even if it is less sensitive to link failures, it is not true that the MLPDF system has better detection performance than the conventional parallel fusion system. In fact we show that the performance results derived in the original paper are based on a wrong assumption and lead to an absurdity.