In order to acquire images on the faster timescales often required by live-cell imaging, laser scanning confocal microscopes must be re-engineered to incorporate advanced scanning scenarios that enable the beam to be raster-scanned across the specimen at higher speeds. To overcome the inherently slow speed of traditional confocal microscopes, several manufacturers have introduced instruments equipped with resonant scanning mirrors that are capable of gathering images at 30 frames per second or higher.
The tutorial initializes by displaying the cosinusoidal pattern of a resonant scanner pixel clock laser beam superimposed on a stretched Ronchi grating (long black lines). Beneath the scan pattern is the light intensity transmitted through the grating and the pixel clock counts generated during the scan. The left-hand side of the diagram shows the forward scan whereas the right-hand side shows the reverse scan. Note that the speed of this tutorial has been dramatically reduced in order to demonstrate the various events that occur during resonant scanning. As scanning proceeds, a typical specimen is illustrated in the Image Scan window. The horizontal scanner mirror angle is also indicated as is the position of the reflected beam from the pixel clock laser with respect to the Ronchi grating. In order to operate the tutorial, click the Scan Image button to enable manual changes to the X Scan and Y Scan sliders.
Due to the non-linearity of resonant scanner velocity, a fluorescent specimen is scanned at the highest speeds in the central region, with the velocity progressively decreasing as the scan reaches the edges. As a result, when the image data flow derived from a resonant scanner is acquired with a frame grabber clocked at a constant pixel rate (which assumes the beam is scanned linearly), the images appear stretched at the edges. Furthermore, the uneven distribution of laser excitation intensity (greater at the edges) produces excessive photobleaching (and potentially phototoxicity) at the edges of the scanned region because of the increased exposure to the laser light. The simplest option for compensating scanning non-linearity is to limit the scanning range to that portion of the oscillation period where the velocity of the galvanometer is almost linear, which occurs over a region spanning approximately 70 percent of the total scan width. Unfortunately, this solution reduces the amount of time that emission signal can be collected, increases the scan turnaround time before signal can be collected again, and doesn't prevent photobleaching in regions of the specimen that fall outside the area being recorded. Photobleaching effects can be minimized, however, when using the linear portion of the galvanometer oscillation by incorporating an adjustable aperture that limits the region of the specimen being exposed to illumination.
Image distortion induced by non-linear resonant galvanometer scanning is fortunately predictable and can be corrected using either software or hardware solutions. Regardless of the correction scheme involved to produce images, the most effective scanning strategy involves collecting data during both the forward and backward scans of the galvanometer mirror. Recording data during the forward scan is straightforward, but due to the fact that the backward scan reverses the direction in which pixels are recorded, the image data must be inverted using specialized read-write buffers or software. In practice, images are gathered at twice the normal width (in effect, 1024 pixels for a final image size of 512 x 512 pixels) and half the normal height (256 pixels). At the end of each forward x-axis sweep of the resonant galvanometer mirror, the vertical sawtooth signal supplied to the linear (y) galvanometer is incremented by one line and the reverse (x) scan commences. In this manner, the vertical scan progresses smoothly for 256 cycles until enough data for one image is acquired.
Both software and hardware clocking solutions for resonant scanning confocal microscopy rely on knowing positional data for the high frequency resonant galvanometer mirror. On 7.9 kilohertz devices, this mirror can rotate approximately 12 degrees in each direction around the central axis (for a total of 24 degrees of rotation). Mirror angular position (θ) and phase or velocity (φ) relationship can be predicted depending upon the state of the oscillating cycle. At the beginning of an oscillation cycle (mirror position equal to plus or minus 12 degrees; see Figure 1) the mirror is stationary with a velocity equal to zero. As the mirror swings toward the midpoint, the angular velocity increases as a cosine function of the phase and reaches its maximum when the mirror angle equals zero. Continuing, the mirror slows down again as it approaches the end of the forward scan until it reaches a velocity equal to zero as it instantaneously reverses direction. Identical changes in angular velocity are observed as the mirror rotates in the opposite direction. Another factor that should be considered with resonant galvanometers is that these devices are high-Q oscillators and cannot be efficiently synchronized to an external frequency due to large response variations in phase and amplitude when even the slightest drift occurs between the natural galvanometer frequency and the external driving source. Therefore, the resonant galvanometer should itself be used as the master oscillator to which all other timing components are synchronized.
Software-based clocking schemes for resonant scanning confocal microscopy involve algorithms that are capable of pixel data linearization using look-up tables based on the predictable motion of the mirror. As an example, one popular algorithm first determines a phase constant for the mirror and frame grabber, and then locates the center pixel of the final image. This algorithm operates around the central plane of symmetry where the mirror position is known with a high degree of precision. Images are processed by examining one horizontal scan line at a time as they are delivered to the frame grabber. Pixel data is stored as received from the scanner at twice the final image width and half the final height. The first step in the process involves inverting data points from the second half of the horizontal scan line, followed by interlacing the inverted data between the data lines for the first half of the image (Figure 1(b)). Because the midpoint of the image may not be known with certainty due to a lag between collection of data and initiation of the horizontal synchronization signal that triggers acquisition, it is often necessary for the software to offset the inverted scan line by a few pixels. In this case, the right edge of the scan line then serves as a reference point for the image. The next step in the sequence is to apply a correction factor for each pixel or phase position relative to the center pixel. That pixel is then relocated to the correct position in the final image (Figure 1(c)). In the ideal case, using sophisticated high speed computers, incoming data can be analyzed on the fly and recorded to the hard drive in real time.
During the horizontal scan period of the resonant galvanometer, the clocking photodiode registers pulses of laser light scattered by the Ronchi grating that can be electronically converted to a pixel clock having variable frequency. These light pulses will be more frequent in time when the galvanometer mirror is in the central portion of its oscillatory period, but they will slow down as the mirror reaches the end of a scan and reverses direction. Among the requirements of the photodiode detector are the ability to handle the variable pulse frequency (in terms of bandwidth) and it must be large enough to detect the entire line sweep. The detector is connected to a transconductance amplifier that transforms photodiode current pulses into an amplified voltage, which is then fed to a frequency doubler circuit that doubles the number of pixels per line. The photodiode and associated electronics deliver 256 pulses per line and the frequency doubling circuit transforms this output into 512 pulses per line that are acquired by a frame grabber and used to construct an image.
Presented in Figure 2 is an illustration of the basic concepts behind the use of a Ronchi grating coupled to a photodiode detector as a variable pixel clock. For purposes of clarity, the grating illustrated in Figure 2(a) contains only eight spacings of equal width superimposed with one complete cycle of cosinusoidal motion from the resonant galvanometer versus time (red curve; 125 microseconds). The actual grating has 256 line spacings and is far more compact in the linear dimensions due to the fact that the pixel clock laser actually retraces a single line (perpendicular to the opaque lines) across the grating surface in operation (Figure 2(b)). As the light beam enters and exits the clear regions of the Ronchi grating, the transmitted light intensity that is detected by the photodiode rises and falls in step with galvanometer motion (Figure 2(c)). Each transition (in this case, from dark to light) of light intensity detected by the photodiode is used to generate the pixel clock (Figure 2(d)) that triggers image capture by the microscope. Although the clock pulses illustrated in Figure 2(d) are non-uniformly spaced in time (ranging from approximately 70 to 160 nanoseconds), they nevertheless correspond to equal intervals in image space.
In order to acquire pixel samples at equal spatial intervals (rather than in equal temporal increments), photomultiplier signals from the primary microscope imaging system pass through conventional amplifiers with variable gain and offset, and are then digitized by an analog-to-digital converter triggered by the pixel clock generated from the Ronchi grating pulses. However, half of the image information is inverted due to the fact that the resonant galvanometer sweeps in both directions. In order to rectify this bidirectional scan artifact before constructing an image, the digital pixel output is first transferred to either a first-in-first-out (FIFO) or last-in-first-out (LIFO), buffer depending upon whether the pixels were acquired from left to right or right to left. Thus, during the horizontal sweep of the galvanometer mirror, image pixels that are acquired during the first half of the sweep (time period of 62.5 microseconds) are fed to the FIFO buffer whereas pixels acquired during the reverse sweep are fed to the LIFO buffer. The FIFO buffer is read and reset to zero at the beginning of each horizontal line acquisition, whereas the LIFO buffer counts up during storage of incoming data and down during the subsequent readout to ensure that pixels acquired during the reverse sweep are inverted.
Jeffrey M. Larson and Stanley A. Schwartz - Nikon Instruments, Inc., 1300 Walt Whitman Road, Melville, New York, 11747.
Adam M. Rainey, Alex B. Coker, and Michael W. Davidson - National High Magnetic Field Laboratory, 1800 East Paul Dirac Dr., The Florida State University, Tallahassee, Florida, 32310.