Precision of light intensity measurement in biological optical microscopy

Abstract

SummaryStandardization and calibration of optical microscopy systems have become an important issue owing to the increasing role of biological imaging in high-content screening technology. The proper interpretation of data from high-content screening imaging experiments requires detailed information about the capabilities of the systems, including their available dynamic range, sensitivity and noise. Currently available techniques for calibration and standardization of digital microscopes commonly used in cell biology laboratories provide an estimation of stability and measurement precision (noise) of an imaging system at a single level of signal intensity. In addition, only the total noise level, not its characteristics (spectrum), is measured. We propose a novel technique for estimation of temporal variability of signal and noise in microscopic imaging. The method requires registration of a time series of images of any stationary biological specimen. The subsequent analysis involves a multi-step process, which separates monotonic, periodic and random components of every pixel intensity change in time. The technique allows simultaneous determination of dark, photonic and multiplicative components of noise present in biological measurements. Consequently, a respective confidence interval (noise level) is obtained for each level of signal. The technique is validated using test sets of biological images with known signal and noise characteristics. The method is also applied to assess

Automatic quality assessment for fluorescence microscopy images

“…The most common source of noise in fluorescence microscopy is the photon detection noise [9], [5], [4], also known as shot noise or intrinsic noise [16]. Another source of noise is the extrinsic noise.…”

“…A model similar to 1 is then fitted to the estimated noise to compute the level of signal dependant noise. These two approaches were also proposed in [5] as two concurrent estimations of the noise.…”

“…We propose a way to estimate the noise in fluorescence microscope images from a single image, taking into account both signal dependent and signal independent noise. Combining both approaches presented in [5], we present a new expression of the signal to noise ratio (SNR), for both global and local estimation. By estimating the SNR locally, we take into account non uniform repartition of the noise as a result of stronger signal (and thus stronger noise) or possibly some detector faults.…”

“…We can then obtain the expression for the calculation of global noise variance seen in [5], by rewriting Eq.7 as age), and we used N9 as an estimate of the global nOIse. We imposed .<4 to get the value of the background noise variance a~g, which means that we neglicted for the fitting the signal-dependant noise in the background.…”

“…In all cases, either gaussian or poisson noise alone is consider. In order to differentiate the noise components and to give their respective levels, Bernas and co-workers [4], [5] fitted a quadratic model of the form…”