Microscopy software speeds cancer research
By R. Winn Hardin,Contributing Editor
A new and important cancer treatment centers on antiangiogenesis or "smart" drugs. These drugs destroy cancerous tumors not surrounding healthy tissue by cutting off their blood supply. In a few years, these drugs are projected to be able to separate cancer cells from healthy cells without the additional complications of radiation-induced organ failure, surgery, sterility, and baldness.
FIGURE 1. To perform drug-testing imaging on rat-retinal tissues for cancer research, VayTek developers are teaming microscopes, PCs, antiblurring algorithms, and automated micropositioning systems to structure automated vision systems that rival laser-scanning confocal microscopes.
Researchers are pursuing a host of these smart drugs and testing their findings on a variety of specimens. Testing antiangiogenesis drugs currently calls for the use of rat retinas. These retinas grow rapidly during development, endure growth factors similar to those of cancerous tumors, and possess complex, networked patterns that closely mirror the arterial structures in humans. However, manual drug testing on retinas is a tedious, time-consuming task.
Measuring the changes in volume of microscopic arteries using 50-µm-thick samples requires precision, fortitude, and repetition that are best handled by automated vision systems.
At the direction of a global pharmaceutical company, VayTek (Fairfield, IA) has developed an automated microscopy system that incorporates microscope and antiblurring algorithms with a confocal adapter to provide data comparable to more-expensive laser-scanning confocal microscopes (see Fig. 1).
Life sciencesTo determine the efficacy of these antiangiogenesis drugs, microscopic changes in the arteries that supply blood to the cancer tumor must be measured. Rat-retina samples are approximately 1 x 1 cm in area and about 50 µm thick. In operation, the microscopy system must take up to 36 images per retina just to image 10% to 20% of the total area of the retina. The system also is required to make three-dimensional (3-D) volume measurements.To accomplish these goals, the microscopy system collects stacks of between 40 and 50 pictures along the z-axis for each of 36 locations to estimate the arterial volume in a 3-D tissue sample. However, there is blur from material above and below the plane of focus. In addition, the cloverleaf shape of the sample mounted to the slide needs special stage-control programming to cover a statistically representative area (see Fig. 2).
Researchers at the pharmaceutical company chose a combination of fluorescent dye and antiquenching agents to prolong the time the dye would fluoresce before losing potency. In operation, a mercury lamp inside the Olympus (Melville, NY) AX70 research system microscope excites the dye that is attached to the arterial walls of a rat's retina. During excitation of the sample, a Peltier-cooled Retiga 1350EX camera from Quantitative Imaging (Burnaby, Canada) attached to the microscope takes upward of 5400 images per slide (see Fig. 3). The exposure time for each image is around 150 ms. With stage and microscope objective movements, total imaging time can take up to 90 minutes per slide.
Confocal benefitsIn a rat's retina, a major artery follows the optic nerve from the brain and stops just short of the top layer of the retina. Smaller arteries radiate out from the main artery, below and parallel to the retinal surface, like spokes on a wheel. This wheel is called the outer plexus. From there, secondary arteries dive back toward the optical nerve for 10 to 20 µm—the middle plexus—before branching into a third layer—the inner plexus. To meet the pharmaceutical company's needs, the VayTek system was required to capture the arterial volume in all three tissue layers.
FIGURE 2. To perform drug inspection and analysis for cancer research, the automated VayTek vision system collects stacks of images from different areas of cloverleaf-shaped rat-retina tissues.
To do so, a series of pictures of the fluorescent retinal tissue was taken at various focal depths along the z-axis. This approach placed heavy emphasis on the optics because the application required tight control of the microscope's focus to collect up to 50 images along the 50-µm z-axis. By isolating and measuring the arteries in the 2-D images, the system software can make measurements of the total arterial volume in the 3-D sample. The solution was a laser-scanning confocal microscope (LSCM).
But, according to Vaytek's president John Kesterson, LSCMs are expensive and can quickly bleach fluorescently tagged specimens. So, Kesterson chose a research-grade fluorescent microscope with an inexpensive confocal adapter.
3-D scanningOriginally developed by Triptar Lens Co. (Rochester, NY) for Optem Avimo Precision Instruments (Fairport, NY), an Optigrid desktop confocal microscope controller gives the AX70 microscope the ability to collect thin, confocal 2-D cross sections of the 3-D retinal sample. A typical imaging session begins at the PC with an operator initiating a batch analysis by defining the pattern of images on each retina, the depth along the z-axis for the image stack, the distance between images in the stack, and the image-acquisition time. A commercial host PC with a Windows-based 800-MHz, Pentium III processor running VayTek's RetinaScan and Triptar's Image Pro driver software controls the Optigrid. Other computer characteristics include 256-Mbyte RAM, a 60-Gbyte hard drive, a 21-in. Sony monitor, and RS-232 and IEEE 1394 (FireWire) ports.
VayTek's software triggers the Retiga camera to capture 1024 x 1024-pixel images. After the first image is captured, RetinaScan software sends a signal to a proprietary Optigrid PCI card in the PC across a special cable that also carries power and data from the Optigrid to the PC.
After the operator selects the image-acquisition time, the Retiga camera captures the first confocal cross section. After the first exposure, RetinaScan software keys a piezoelectric motor inside the Optigrid and triggers the Retiga camera again. The process repeats until three images reside in the PC's RAM. Image processing of these three images results in the final desired image.
Cancer's deathAfter the Optigrid driver supplies the finished image to RetinaScan software processing, the VayTek software sends stage-control commands via an RS-232 cable to the Ludl (Hawthorne, NY) MAC 5000 stage control box. This box separately sends electrical control signals across 15-pin cable connectors to the x, y, and z stages. RetinaScan software also sends command controls for objective, lighting, or focus changes across a second RS-232 connection to the Olympus AX70 microscope.After a stack of images has been collected, RetinaScan software analyzes the images to extract the fluorescent signals from the dyed arteries and then make volumetric measurements. According to Kesterson, VayTek uses Media Cybernetics (Silver Spring, MD) Image Pro Plus software as a library for some of the low-level image-processing tasks. VayTek's developers then customized the RetinaScan software using Visual Basic and C++ algorithms to enable a program that can isolate the arteries, measure their volumes, and determine their locations.
FIGURE 4. VayTek's RetinaScan software leverages the familiarity of Microsoft's Excel spreadsheet by tying it to optical cross sections of the retina. By moving the computer mouse pointer along the z-axis image, the operator can automatically shuffle through the images that created the volume.
RetinaScan takes the volume measurements and cross-references them against 2-D image and 3-D stack locations in an Excel-based spreadsheet. The system automatically graphs the outputs expressed as a volume versus z-axis position. Then, the operator can automatically shift between spreadsheet/graphic representations and image data by scanning the computer-mouse pointer along the line graph. RetinaScan's reporting functions automatically shuffle through the images that formed the basis for that point on the graph. From there, the operator can save, print, or make manual corrections to the data.
By inspecting the graphed data and the images in the stacks simultaneously, the operator can quickly verify the accuracy of the analysis. The spreadsheet data can be further analyzed to determine the effects of a drug treatment on the growth patterns of the arteries. The fast and efficient visual review of several hundred images can also reveal anomalies in the growth patterns that the statistical analyses might miss (see Fig. 4).
The system project took less than 12 weeks to integrate, says Kesterson, and most of the time was spent creating the volume-extraction algorithms and fine-tuning the acquisition routines as a result of the volume algorithm development. VayTek's software offered command controls for all automated microscopes, stages, and most microscope cameras, so little time was spent tailoring the software to the hardware.
According to Kesterson, developing the RetinaScan software was more than just an opportunity to develop a solution to a difficult imaging problem. It is the first of several turnkey systems that VayTek plans to offer for high-volume throughput imaging and inspection of 3-D samples. The combination of the low-cost Optigrid unit and the Retiga camera results in a system that can make volume measurements of a large number of specimens. VayTek has begun another project to incorporate the scanning of 96 and 384 well plates, as well as microdot arrays on slides.
Company InformationLudl Electronic ProductsHawthorne, NY 10532Web: www.ludl.comMedia Cybernetics
Silver Spring, MD 20910
Web: www.mediacy.com
Olympus America
Melville, NY 11747
Web: www.olympusamerica.com
Optem/Avimo Precision Instruments
Fairport, NY 14450
Web: www.optemintl.com
Quantitative Imaging
Burnaby, BC, V5A 1W9 Canada
Web: www.qimaging.com
Sony Electronics Inc.
Park Ridge, NJ 07656
Web: www.sony.com
3COM
Santa Clara, CA 95052
Web: www.3com.com
Triptar Lens Co.
Rochester, NY 14607
Web: www.triptar.com
VayTek
Fairfield, IA 52556
Web: www.vaytek.com