Fig 1.
Illustrating the signal-to-noise ratio model of an EMCCD camera.
(A) The path of photons (green) emitted by the light source, arriving at the camera sensor, get converted into electrons (-) that are amplified and converted into pixel intensity by the Analogue-to-Digital converter (ADC). (B) The raw intensity at high and low signal to noise ratios showing the signal of interest relative to the noise.
Fig 2.
Verifying the noise model and measuring camera parameters.
(A) Mean standard deviation in pixel intensity of five 0-gain, 0-s exposure dark images (light blue) and their associated difference images. (B) Dark current noise at different exposure times (0 s – 600 s). The ‘marketed’ dark current noise at different exposure times was computed based on the market dark current of 1 e-/s/pixel, using Eq. 11. The ‘observed’ dark current noise was determined from images captured with a closed camera shutter, using Eq. 10. If the dark current cannot be detected (), then the dark current noise is set to a default value of 0. The ‘fitted’ dark current noise was computed based on the fitted dark current of 2.5 e-/s/pixel. (C) Photon shot noise of images at various exposure times (20 ms – 80 ms) with light and an open camera shutter (observed). The observed photon shot noise was calculated by removing dark image noise from the image noise using Eq. 13. The expected photon shot noise was calculated based on the observed photon shot noise at the smallest exposure time, using Eq. 14. (D) Standard deviation in pixel intensity of dark images at three different electron multiplication gains (2.40, 9.03 &103.5) on a log-10 scale (observed). The ‘literature reported’ dark image noise was calculated based on the previously reported CIC value of 4 e-, using Eq. 15. The ‘fitted’ dark image noise was calculated based on the fitted CIC value of 25 e-.
Table 1.
The photon shot noise, dark current noise, clock-induced charge (CIC), and readout noise of an EMCCD camera for microscopy.
Fig 3.
Effect of an additional emission filter on signal-to-noise ratio.
(A, F) Comparing the signal to noise ratio between the OEM filter setting and additional emission filter (Double Em) for EGFP (A) and mCherry (F) channels. OEM stands for the original equipment manufacturer’s microscope setup. Plots 1 and 2 show the background intensity and excess background noise relative to the dark image for images with no agar pad or oil (No agar), agar pad + oil but no cells (No cells), agar pad + oil+cells (With cells). Plots 3 and 4 show background subtracted cell intensity at 3 locations on an agar pad (4 for mCherry) and relative change in camera independent SNR of images with agar pad + oil+cells. SNR improvement is only meaningful if the improved setting shows similar or lower background subtracted cell intensity, but much lower relative excess background noise. (B) Schematic of the path of light from the fluorescence illumination system to the detector in an inverted widefield microscope using the manufacturer’s recommended setup. Created in BioRender. (C,D) Absorbance vs. wavelength spectra of EGFP single emission and excitation filters showing escaped excitation light (blue arrow) for C and blocking residual excitation light by the second emission filter for D. The width of each arrow-stem qualitatively indicates the amount of light. Created in BioRender. Note the position of the detector and light source do not reflect their true location in the microscope. (E) Absorbance vs. wavelength spectra of mCherry single excitation filter but double emission filters showing blocking of residual excitation light but also blocking of emission light past 625 nm.
Fig 4.
Effect of wait time or low bright field light on signal-to-noise ratio.
(A) Above dark image background intensity and photon shot noise of the background for images captured at various wait times (BFP: 0 s – 12 s, EGFP: 0 s & 6 s) in darkness after bright field illumination (high and low intensity) but before fluorescence acquisition. (B) Above dark image background intensity and relative excess background noise for images taken during the transient darkness prior to fluorescence acquisition and during BFP fluorescence acquisition. (C) 0 s vs. 6 s wait time for BFP acquisition of E.coli cells under weak rhamnose induction of BFP. Plots 1 and 2 show above dark image background intensity and relative excess background noise for images taken at 3 different agar pad locations. Plots 3 and 4 show background subtracted cell intensity and relative change in camera independent SNR at 3 agar pad locations. SNR improvement is only meaningful if the improved setting shows similar or lower background subtracted cell intensity, but much lower relative excess background noise.
Fig 5.
Effect of an additional excitation filter on signal to noise ratio.
(A) Absorbance vs. wavelength spectra for a hypothetical scenario where two emission filters are blocking escaped excitation light and a single excitation filter is transmitting residual emission wavelength (green) from the illumination lamp. (B) Absorbance vs. wavelength spectra for a hypothetical scenario where double emission and double excitation filters blocking escaped excitation light and emission wavelength from the illumination lamp, respectively. (C) The current model and new model of the path of light in a standard inverted widefield microscope setup. The new model shows escaped excitation light that reaches the detector, emission light that surpasses the single excitation filter, and reflected bright field light of the optic system. (D,E) EGFP acquisition using the OEM setup vs. double emission filters vs. double emission + double excitation filters. The images acquired under different microscope settings were processed to make them comparable: for each image, pixel intensity was subtracted by the mean pixel intensity of the background region in that image. The magnitude and amount of static in a background region reflects the background noise level. Plots 1 and 2 in panel E shows background intensity and excess background noise for the dark image, agar + oil but no cell images, and agar + oil + cell images. Plots 3 and 4 in panel E shows background subtracted cell intensity and mean camera independent SNR of images at 3 different agar pad locations. SNR improvement is only meaningful if the improved setting shows similar or lower background subtracted cell intensity, but much lower relative excess background noise.