It is useful to have quantitative measures of image quality, so that you can systematically determine the effect of changing different imaging conditions on experiment results. In the last post we showed that averaging can improve image quality when carrying out high-speed imaging. Qualitatively, the averaged images looked brighter and were more clearly defined (ie: they had more contrast). Yet visual inspection will only get you so far. Instead, it’s important to use quantitative metrics when gauging the effect of averaging or other imaging parameters on the quality of data sets. A quantitative metric also helps with consistency, reproducibility and trouble-shooting.
Image quality can be evaluated by several approaches that are generally grouped together as signal-to-noise measurements. A signal-to-noise measurement accounts for all sources of noise in an imaging experiment. However, rigorous and thorough signal to noise measurements can be complicated and time intensive. For this reason in bioimaging, a simplified metric known as signal to background (StB) is often used instead. Although StB measurements are not as rigorous as signal to noise determinations, StB often gives us enough information for routine bioimaging.
In a common approach, the StB measurement divides the average signal intensity of the object (the sample) by the standard deviation of the signal intensity in the image background. Fig.1 shows a fluorescence image of a pollen grain sample overlaid with several regions of interest. In yellow, you see the regions around the pollen grains used to calculate the average signal. The cyan rectangles are the regions of interest used to calculate the background average intensity and its standard deviation.
The average signal intensity and standard deviations within each ROI can be calculated and analyzed using a program such as FIJI or using commercial or custom software. Using this method, you can compare the StB values for different experimental conditions. This allows you to explore to which extent different experimental conditions affect your image quality.
Fig.1 A pollen grain sample imaged using high-speed multiphoton microscope (Bliq VMS). The yellow ROIs surrounding the pollen grains were analyzed to give the mean intensity of the signal. The rectangular cyan ROIs were selected to represent the background; the standard deviation of the background intensity values were calculated. All measurements were done in Fiji.
Table 1 shows the StB obtained for the pollen grains using two different scanning speeds and three different averaging conditions (1x – no average, 4x, and 8x).
Table 1. No. averages: number of frames acquired and averaged for final image; Mean intensity: mean value of the signal (yellow ROIs around the pollen grains in Fig.1), StDev: Standard Deviation of the signal intensities in the background (cyan rectangle ROIs in Fig.1), StB: signal-to-background ratio for each condition (ie: number of averages).
As can be seen in the StB column in Table 1, the signal-to-background improves with increasing averaging. In this application, we stopped at 32x even though it’s probable to obtain further improvement in quality. Although the StB here is trending higher, increasing the number of averages requires more time for image acquisition, and may risk photobleaching the sample. For live cells, averaging for longer times might also cause phototoxicity and blurring if the cells move during the acquisition. This highlights a central issue in imaging- although it’s often possible to improve an image, often there is a cost to pay for the desired improvement. In sum, when testing the effects of different imaging conditions, it’s also important to apply the StB method judiciously, when balancing the sometimes competing needs of an imaging experiment.