Abstract

Confocal scanning is an effective modality for three-dimensional microscopy because of its optical sectioning capability. However, there is a tradeoff to be made between resolution and signal-to-noise ratio (SNR) when fluorescent objects are viewed by confocal scanning. Under conditions of limited photon dose to the specimen, enlarging the confocal detector aperture admits more light and improves the SNR at the expense of resolution. Conversely, making the aperture smaller improves the depth resolution in principle, but it can also reduce SNR to the point where closely spaced elements in the object are effectively unresolvable by a human observer. We present in this paper an experimental study using psychophysical ROC analysis and the Hotelling trace criterion to evaluate resolution in this partially confocal regime. Synthetic images of objects at various confocal aperture sizes and noise levels are shown to human observers who are asked to perform a classification task requiring depth resolution. The collected responses are summarized by ROC measures. We compare these with the results of analyzing the same image sets with the Hotelling trace criterion, which has been observed to track human observer performance in certain radiological tasks.¹

© 1992 Optical Society of America