Statistics of Particle Detection From Single-Channel Fluorescence Signals for Flow Cytometric Applications

Abstract

Flow cytometers and other particle analyzing devices require the reliable detection of particles for counting, classification, sorting, and other applications. Using miniaturized microfluidic devices opens up new application possibilities, but also poses new challenges for reliable signal detection. Simplifying flow cytometers by eliminating scattering as the second signal channel changes the statistics for fluorescence particle detection due to the now unknown number and position of particles in the fluorescence signal. Furthermore, in applications such as particle sorters, the detection delay must be kept short, and cost considerations may limit the available signal processing power. In this paper, we investigate the statistics of a simple thresholded peak detection algorithm with a defined spatial resolution that satisfies these needs. We define a signal system with statistical properties in accordance with measurements obtained from a miniaturized fluorescence particle detection device. For this model, we derive properties, such as spatial resolution, rate of particle interference, and false positives and false negatives due to noise. Using these properties allows us to predict the device performance during the design phase and therefore optimizes the particle detection device for specific applications to reduce device cost, while still maintaining crucial performance figures. In order to evaluate the applicability of the signal model in a real application, we compare the theoretical amplitude distributions with experimental results.

Particle detection from spatially modulated fluorescence signals

“…Particle detection from local maxima as used in Ref. 25 is not feasible with such a signal. In order to suppress the side maxima, special selection rules have to be employed but the particles may not be very close or else they start generating additional maxima as high as the main maximum and false positives will be detected.…”

“…The ratio of the height of the main maximum to the height of the side maxima reaches values of up to 5 for the DCS filter which is consistently higher than for the matched filter but still does not allow a dynamic range of over 100 obtained by non-modulated detection. 25 Considering the results in Ref. 25 of a non-modulated system where particles of vastly different amplitudes can be placed very close to the resolution limit without causing false positives, these side maxima are a serious problem.…”

“…25 Considering the results in Ref. 25 of a non-modulated system where particles of vastly different amplitudes can be placed very close to the resolution limit without causing false positives, these side maxima are a serious problem. We therefore analyze the system and devise a method for the filter calculation which provides a good compromise between PSNR and suppression of side maxima.…”

“…The pulse form is not the smooth shape with a single maximum like in Ref. 25. Instead it contains many local maxima due to the sequence of open and closed subapertures used.…”

Fluorescence particle detection using microfluidics and planar optoelectronic elements

“…12 and analyzed the signal statistics in Ref. 13. Furthermore we have analyzed the influence of the noise sources on the coefficient of variation (CV) in Ref.…”