Image registration and averaging of low laser power two-photon fluorescence images of mouse retina

Abstract

Two-photon fluorescence microscopy (TPM) is now being used routinely to image live cells for extended periods deep within tissues, including the retina and other structures within the eye . However, very low laser power is a requirement to obtain TPM images of the retina safely. Unfortunately, a reduction in laser power also reduces the signal-to-noise ratio of collected images, making it difficult to visualize structural details. Here, image registration and averaging methods applied to TPM images of the eye in living animals (without the need for auxiliary hardware) demonstrate the structural information obtained with laser power down to 1 mW. Image registration provided between 1.4% and 13.0% improvement in image quality compared to averaging images without registrations when using a high-fluorescence template, and between 0.2% and 12.0% when employing the average of collected images as the template. Also, a diminishing return on image quality when more images were used to obtain the averaged image is shown. This work provides a foundation for obtaining informative TPM images with laser powers of 1 mW, compared to previous levels for imaging mice ranging between 6.3 mW [Palczewska G., Nat Med.20, 785 (2014) Sharma R., Biomed. Opt. Express4, 1285 (2013)].

A Poisson-Gaussian Denoising Dataset With Real Fluorescence Microscopy Images

“…The results are in good agreement with the theory in Equation 7 and the observations in Table 1, as the estimated parameters follow the trend of a/S and b/S, and their initial values (S = 1) are close to the ones in Table 1. Figure 4 also shows that the values of a and b exhibit little change when the number of captures used for averaging is more than 40; this confirms the observation reported in [3] that the image quality or noise characteristics of a fluorescence microscopy image will see little improvement after including around 40 images in the average.…”

“…We make S = 1, 2, 4, 8, 16 to create images with five different noise levels, and S = 50 to generate the ground truth. As demonstrated in [3] and also shown in Section 4.3, for fluorescence microscopy images, little image quality improvement can be seen after including around 40 images in averaging.…”

“…The main efforts are spent on image pre-processing, such as image registration to remove the spatial misalignment of an image sequence with the same field of view (FOV) [3,1], intensity scaling due to the changes of light strength or analog gain [24], and methods to cope with clipped pixels due to over exposure or low-light conditions [4]. The images captured by commercial microscopes in our dataset turns out to be well aligned, and the analog gain is carefully chosen to avoid clipping and to utilize the full dynamic range.…”

“…Ground truth estimation We estimate the ground truth by averaging all 50 captures on the same FOV, similar to the approaches employed in [3] and [17]; hence in the FMD dataset, each FOV has only one ground truth that is shared by all noise realizations from that FOV. As demonstrated in [3] and also shown in Section 4.3, the image quality or noise characteristics of a fluorescence microscopy image will see little improvement after including around 40 images in the average; therefore, we choose 50 captures as our criterion to obtain the ground truth. As shown in Equations (6) and (7), the ground truth y i for images with different noise levels z j i is the same, and image averaging is equivalent to sampling from a Poisson-Gaussian distribution with a higher SNR.…”

Vision science and adaptive optics, the state of the field

“…GP: To excite specific endogeneous and artificial fluorophores in the retina (Alexander et al, 2016), we apply tunable lasers with wavelengths appropriate for two-photon excitation of the intermediates and by-products of the retinoid cycle. The two-photon excited fluorescence is then collected in a non-descanned manner, to maximize the signal.…”