Confocal microscopy is often used to make three-dimensional high-resolution images of biological samples. It is done by reflecting a laser beam off of a dichroic mirror, and focusing it on a sample with the use of a microscope. This sample is treated with a fluorescent dye. The dye molecules emit light of a higher wavelength with an intensity proportional to their illumination intensity. This light travels back, and can now pass through the dichroic mirror. Hereafter, the light is focused onto a pinhole that blocks out-of-focus light. By measuring the intensity of the in-focus light, and raster scanning the focused light spot across the sample, a high resolution image of a section of the sample can be formed. The raster scanning is done with a set of rotating scan mirrors. To attain a three-dimensional image, the spot can be focused on a new section of the sample, and the process is repeated.
Rescan confocal microscopy is a relatively new addition to it. Instead of directly measuring the light intensity after it passes through the dichroic mirror, it is instead recollimated and rescanned across a CMOS chip with a second set of rotating rescan mirrors, synchronized with the scan mirrors. The rescan surface area is proportionally larger, resulting in a higher resolution than conventional confocal microscopy. The pinhole can also be made larger, resulting in higher quantum efficiency.
State-of-the-art rescan confocal microscopy utilizes two sets of galvanometer mirrors for both the scanning and rescanning, and is very slow; only up to 2 frames per second for a 512 by 512 pixel image. The goal of this research is to speed up a rescan confocal microscopy significantly from the state-of-the-art.
The raster scanning and rescanning is done much like reading a book. In each set there is a fast-axis mirror travelling from left to right, and a slow-axis mirror only moving down a small step.
Thus, the fast-axis mirrors for both the scanning and rescanning bottlenecks the speed. Replacing these two with one faster mirror-based alternative would improve the system speed. Using one mirror-based alternative eliminates potential synchronization issues between scanning and rescanning at high rates.
The three mirror-based alternatives selected for further investigation are: polygon scanners, resonant scanners, and MEMS resonant scanners.
Polygon scanners have a polygonal head with multiple mirror facets. This head is attached to a shaft which is rotated in the order of several thousands of revolutions per minute. Each passing mirror facet equates to one scan line.
Resonant scanners contain a mirror attached to a shaft. This shaft is brought up to torsional eigenfrequency (in the order of several thousands of kilohertz), resulting in the mirror oscillating back-and-forth. Each oscillation period can thus cover two scan lines. MEMS resonant scanners operate on the same principle as conventional resonant scanners, their main difference being their production method and form factor.
The three scanner types were further investigated for and compared to one-another. The performance metric that was used for this was the scanned area per second of the microscope image plane.
Optimization models were made for the three scanners, with this metric and found constraints. The number of variables for all models was reduced to two so that contour plots could be made. These contour plots were made for both the theoretical maximum performance of each scanner, as well as the current commercially-available products. These models show that MEMS resonant scanners have the potential for the largest scanned area per second with up to 2800mm^2/s. However, this is a theoretical large diameter MEMS scanner (2.46mm). While there is recent research focusing on large diameter MEMS resonant scanners, most scanners found in literature are between 1-2mm. Thus, there is most certainly future potential in large diameter MEMS resonant scanners for confocal microscopy applications. However, when comparing the current lineup of
commercially-available scanners MEMS scanners are not feasible due to their small mirror diameters. Polygon scanners attain the highest performance here, with a up to 1100mm^2/s. Conventional resonant scanners can only provide up to 698mm^2/s. Thus, a polygon scanner was chosen for the final design.
Polygon scanners have facet-to-facet errors. One such error: the cross-scan error, or dynamic track error could form a problem for the final image and result in brightness banding, due to the (re)scan lines jumping up or down along the slow-axis with each passing facet.
It was decided that this error had to be eliminated. This can be done by using a relay and a multiple of two mirrors. A relay relays and inverts an angle four focal lengths further along the optical axis. Thus, by relaying the cross-scan error back to the same polygon facet, the error in the slow-axis can be completely eliminated while maintaining the desired angle along the fast-axis. This method appears to be novel in literature, and was subsequently verified with an experiment.
This novel solution was used in the final polygon design, which was fully detailed in CAD with off-the-shelf optical components. The made Zemax models showed a diffraction-limited system. The final area scan rate is 949mm^2/s, and is 42 times faster than the state-of-the-art. Future work includes assembling and testing the prototype setup, and converting it into a commercial product.