For decades, scientists have endeavored to convert visual microscope images into digital data for storage, computation, and viewing. Reproducing the quality of images seen directly through a microscope is a particularly difficult challenge. Currently, researchers are using CCD based cameras added to traditional microscopes to meet this need.
COSMIC uses an entirely different imaging technology to produce high-resolution, full color images at frame rates fast enough to examine live samples. In fact, it does not contain a camera at all, but uses a patented spot scanning technique that has many advantages over conventional CCD cameras. The images produced are limited in resolution only by the microscope optics and not by the detector resolution. The system uses the same photomultiplier tube (PMT) detectors as in Confocal and MultiPhoton fluorescence microscopes, taking advantage of the low light and high-speed properties inherent in these devices. COSMIC stands for COlor Scanning MICroscope.
The best digital cameras placed on high-end fluorescence microscopes have significant shortcomings. As the number of pixels increase on newer CCD detectors, the sensitivity goes down, requiring either higher illumination, or longer exposure times, or both, which can be damaging to delicate samples. The typical fluorescence illumination system consists of a 100W Hg lamp that floods the sample with so much radiation that eye shields are used to protect the operator. Photobleaching and cell death are major problems, giving short examination times, and requiring preparation of extra sample material for multiple runs.
The regular array of pixels on the CCD detector are subject to aliasing patterns caused by interference with the sample image. Saturation of the detector with light can cause blooming effects. Relative to brightfield systems, fluorescence systems have higher equipment costs, higher maintenance costs, and greater expense in operating a fluorescence facility.
The COSMIC technology is being applied to these problems to create a revolutionary new digital imaging microscope that will not only combine brightfield, fluorescence, automation, and communication into a single package, but it will operate at the highest resolution, with low impact on the sample, significant ease of use, and with advanced imaging features not available on traditional digital camera systems. The system has none of the problems of CCD detectors. Unlike laser systems, it operates in full color at real time frame rates, making it much more universal in application. A high-powered system with a lamp life of perhaps 200 hours and a cooled CCD camera are expensive parts of a system. The equivalent COSMIC parts cost much less and have a lifetime over 10,000 hours, making COSMIC much better suited to imbedded OEM applications.
THE COSMIC TECHNOLOGY
In order to understand how COSMIC is different from CCD imaging devices and laser scanners, the following section describes the current technology and its features.
COSMIC is based on a flying spot scanner system using a special cathode ray tube (CRT). The CRT spot with its raster scan pattern is imaged through an optical system and focused onto the sample by the objective lens (Figure 1). The spot size on the sample is the smallest that can be created by the objective, called the Airy disk, so the system resolution is limited by the resolution of the objective, not the detectors. The spot illuminates only one point on the sample at a time, reducing bleaching and scattered light. Light transmitted through the sample is divided into RGB spectral ranges and detected by three photomultiplier tubes (PMTs).
The amplified signal is processed by a digital imaging board and then sent real time to a PC for
display, analysis, communications, or storage. From this fundamental image creation technology comes a number of unique features.
True Color - Each pixel in the 1280 x 1024 field contains the full source color spectrum, and is digitized into 24 bits of RGB color, rivaling direct observation through a conventional microscope. For a comparison of COSMIC and CCD cameras, see Appendix 1.
Speed - The frame rate is 13.4 Hz, for full 1280 x 1024 x 24 bit images, which is fast enough to examine live cells in suspension. The display shows the images at 60-72 Hz to minimize flicker.
3x Zoom - The raster scan size can be reduced instantaneously up to 3x, causing a 3x increase in image size on the display. Brightness, focus, and the 1280 x 1024 pixel field size are unchanged, and there is no pixelation of the image. Figure 2 illustrates the effect.
Spot modulation - The scanning spot brightness can be modulated for every pixel in the image to enhance or suppress contrast. For enhancement, the spot goes white where the sample is white and dark where the sample is dark. For suppression the spot goes white where the sample is dark and dark where the sample is white. The function operates at full frame rate, using no computational resources. In Figure 2, contrast suppression was used to show more detail in the dark areas.
Super-resolution - The diffraction limited scanning spot can reveal features smaller than classical optics theory predicts.. As the Zoom feature decreases the scan area, each Airy disk on the sample is digitized into more and more electronic pixels in a process called oversampling. There is a limit beyond which the resolution does not improve and empty magnification results. However, features 1/10 the size predicted by classical theory have been imaged. See Figure 3.
Illumination - The power on the sample in brightfield mode is less than 100 nW. In fluorescence mode, the power on the sample is less than 1 nW. The phosphor light source does not change color temperature with brightness as occurs when a lamp is turned up and down. The CRT has a lifetime of about 10,000 hours, as opposed to about 200 hours for a 100W fluorescent Hg lamp.
A set of digital software features adds further functionality including,
Most digital imaging microscopes are assembled by the buyer from various components, including a microscope, lamp system, CCD camera, frame grabber, computer, monitor, and software (Figure 4). A typical fluorescence illumination system includes of a 100 Watt Hg lamp that floods the sample with so much illumination that eye shields are used to protect the operator. Photobleaching and cell death are major problems, giving short examination times, and requiring preparation of extra sample material for multiple runs. Laser systems are also known for high intensity illumination in the milliWatt range that causes photodamage to samples. Workarounds include synchronized shutters to block the light, adding cost and complexity to the system.
Television format CCD cameras are the most popular imaging devices. However, television camera resolution is much less than direct viewing through a conventional microscope and is generally not clinically accepted for diagnostic quality images. The highest-resolution digital cameras have lower sensitivity, slower frame rates, and are best suited for taking high quality still images, in a fashion similar to photography. The fundamental characteristics of CCD cameras include:
CCD detectors. CCD cameras come in three types: single chip monochrome, single chip color, and triple chip color. Popular single chip color arrays have an alternating pattern of red, green, and blue pixels which are interpolated to create RGB data for each pixel position. COSMIC detectors sense only brightness; they have no pixels, and all data points are real, noninterpolated RGB.
Sensitivity. As manufacturers decrease CCD pixel size to obtain higher resolution, the sensitivity also decreases. Three chip cameras are the least sensitive due to the filter system that creates the three color channels. Solutions to sensitivity problems include increasing the light intensity with high powered lamps or lasers and cooling the CCD detector to reduce the dark noise and increase sensitivity. The uncooled PMT detectors in COSMIC have 1,000 to 10,000 times more sensitivity than CCDs, creating color images with one million times less power on the sample. New PMT type detectors have been introduced with gains exceeding 10^8.
Aliasing. If the sample has a periodic structure whose spacing is close to the CCD detector array spacing, a beat frequency can be generated resulting in wavelike interference patterns, called aliases, or Moiré patterns, that are not real in the image (see Figure 5). The effect is worst at low power magnification. COSMIC has no detector array to cause aliasing. In Figure 5, the pattern is created by crossing two diatoms, whose fine structure causes the beat pattern.
Blooming. When a CCD detector pixel is saturated with light, excess light is scattered into adjacent pixels, causing a star pattern, or blooming in the image. Many cameras have special circuits to minimize blooming, which temporarily halt the operation of the camera. The independent pixels on COSMIC cannot exhibit blooming.
Full field illumination. In an ordinary microscope, all points in the image field are illuminated simultaneously. The scattered light can reduce contrast and contribute to blooming. The spot illumination of COSMIC results in very high contrast, which can eliminate the need for staining on some samples.
No Zoom. CCD cameras have no equivalent to COSMIC’s zoom, or spot modulation features. Zooming is typically done by expanding the pixels in the image, causing visible pixelation.
Operating Expense. Fluorescence imaging components cost much more than equivalent COSMIC imaging components and have shorter service lifetimes. High power lamps have high heat dissipation and lifetimes 50 times shorter than COSMIC. In addition, COSMIC components have no heating problems and can be located internal to OEM systems.
New Transmitted Light Fluorescence
Another unique feature of the COSMIC instrument is transmitted light fluorescence. The earliest fluorescence microscopes were made on transmitted light stands with a straight-through optical system. Poor performance, including leakage of the excitation through the emission filters, caused high background signal and poor contrast. This method was abandoned in favor of epi-fluorescence which illuminates and images through one objective lens. Epi-fluorescence uses an additional dichroic beamsplitter to further separate excitation and emission optical paths, thereby reducing the background light leakage. Advances in filter and beamsplitter technologies continue to improve performance.
Epi-fluorescence has been the standard technique for so long that most scientists know nothing else. In recent years, confocal and multiphoton microscopes using the epi-illumination method have become quite popular for fluorescence, but at very high prices.
Today, advances in filter technology allow a return to transmitted light fluorescence (see Transmitted light fluorescence microscopy revisited, P T Tran; Fred Chang, 10/01/2001, The Biological Bulletin, pg 235, Vol 201, Issue 2). Elimination of the dichroic beamsplitter results in higher optical efficiency and lower cost.
However, a more important benefit comes from the separation of the excitation lens (the objective), from the emission collection lens (the condenser). Figure 6 illustrates the design. With epi-illumination, the objective is both the illumination lens and the collection lens. Because low power objectives have low numerical apertures, they make poor collectors for weak fluorescence radiation.
For example, a 2.5x objective lens with a 0.075 numerical aperture is a very inefficient collector lens. A typical 0.90 NA condenser lens provides 144 times the collecting power for fluorescent emission. One could even use a 1.40 NA oil condenser with an improvement of 348 times. Therefore, transmitted light fluorescence offers much higher collection efficiency for low magnification powers. Low magnification fluorescence with a large field of view could open new methods for the examination of embryos with fluorescence.
Another advantage of transmitted light fluorescence is the distance the imaging light travels inside a sample before it leaves and is collected. In an epi-illumination system the excitation light travels into the sample to excite fluorescence in the focal plane. The emitted fluorescence light then must travel back out of the sample to be imaged by the objective. This is called a two-pass system; the light must travel twice though the same distance. The total optical path varies from 0 for the top surface to (2 x Sample Thickness) for the bottom surface.
In the COSMIC transmitted light method, the excitation light travels into the sample, and the emitted light travels out of the sample on the opposite side. The total light path distance is always the thickness of the sample (1 x Sample Thickness). This is a one pass system. Cosmic can look deeply into a sample and show a clear image (see Figure 7).
The concept that a fluorescent phosphor could have enough energy to excite other fluorescent compounds appears to conflict with reason. However, the original brightfield COSMIC system could always detect bright fluorescence, such as shown in Figure 8. The image is the autofluorescence of a 10µm thick mouse tissue section showing hair follicles, taken with a 10x objective. Excitation was at 525+/-25nm and emission at 610+/-20nm. The built-in averaging and summing modes were used to add gain and reduce noise in the image. The total image exposure time was about 10 seconds.
The purpose of the COSMIC NanoWatt fluorescence project is to extend the sensitivity of the COSMIC system down to very weak fluorescence samples and reduce exposure times.
NanoWatt™ Project Prototype Results
The new prototype COSMIC™ instrument is shown in Figure 9.
Measurements of the energy incident on a sample for a single filter (535nm) and different objectives are shown in Table 1. The minimum power measured was 100 picoWatts, which is near the lower limit of the measuring instrument.
Test samples included calibration beads and biological samples.
The fluorescence was calibrated with Beckman-Coulter ImmunoBrite Beads used for linearity measurement in flow cytometers. The bead diameters are all about 10µm and are rated for excitation at 488nm with emission from 500-700nm. The beads come in four intensities and one set of dummy beads with no dye. The intensities range from bright to very dim on a conventional fluorescence microscope.
In order to use the beads effectively on COSMIC, they were calibrated on a linear scale at the University of NC at Chapel Hill using a conventional fluorescence microscope with a CCD camera. The same setup was used for all beads and the integration times were used as the brightness reference. The normalized times ranged from 1 (brightest) to 3000 (dimmest). According to the UNC lab director, most biological samples are equivalent to the 1200 brightness level beads.
The calibrated bead testing was done with a 63x oil objective / 1.4na oil condenser pair. Images of the dimmest beads, rated at the lowest brightness (3000) are shown in Figure 10. Figure 10 (a) is the brightfield image. Figure 10(b) is an RGB composite of the (c), (d), and (e) images, where filter set “DAPI” is mapped into the blue, set “FITC” is mapped into the green, and filter set “Rhodamine” is mapped into the red channel. Two beads in the (a) image do not appear in the other images, indicating they are dummy beads. The assignment of the colors is arbitrary.
Figure 10. Images of the lowest brightness ImmunoBrite beads. (a) brightfield image, (b) RGB composite of fluorescence images, (c) “Rhodamine” Filter set, (d) “FITC” filter set, (e) “DAPI” filter set.