The Automated Pathologist
Vikram Sheel Kumar, Molly Webster · Clinical Chemistry · 2013
Machines are our friends—they let us quickly deposit a check, zip through a highway toll, or check into a flight—until they take our jobs. The jobs that go first are those that can be automated with simple logic. In the clinical laboratory, machines move specimens into other machines that mix and measure to give results. The pathologist has remained at the top of the ladder, secure by her years of experience and ability to deftly scan a slide and pick out a suspicious-looking cell. No machine could do that. Or could it? We spoke to Dr. Alan Nelson, founder and CEO of VisionGate, Inc., and professor of bioengineering and physics at Arizona State University, who believes that his latest invention could “render microscopy using morphology stain obsolete.” Nelson is no newcomer to automating pathology. Nearly 20 years ago, he took public a company called NeoPath (since acquired by Becton Dickinson) that he founded around a technology he invented to automate Pap smear screening (now available under the name FocalPoint). The Cell-CT™ is a 3-dimensional (3D) cell-imaging platform that Nelson created after observing the limitations of conventional 2D microscopy during his work with NeoPath. The first tests of the Cell-CT platform have been with sputum samples in developing an early screen for lung cancer. “I thought the improvements that 3D might offer would overcome certain issues that have discouraged the development of lung cancer screening from a noninvasive sputum sample,” he shared. “Such as poor diagnostic accuracy by cytopathologists due to the fact that actual cancer cells in sputum can be prohibitively rare for human analysis.” The Cell-CT is a stand-alone in vitro diagnostic device 4 feet high with a base 2.5 feet square. It processes any cell suspension, with the current focus being on blood and sputum samples. The samples are prepared manually with conventional fixation and staining methods, although, according to Nelson, an automated sample-preparation device should become available in the future. With the Cell-CT, a sample is injected under pressure into a disposable cartridge that contains a rotating microcapillary tube. A rotating and translating charge-coupled device camera scanning this tube collects 500 images every 0.72 degrees of tube rotation (Fig. 1). The individual images are used to generate a 3D tomographic image of a single cell that the system may analyze for classification. Similar to flow cytometry, cells are injected into a capillary tube; however, the tube in the Cell-CT is unique because it rotates axially as cells flow through it. As the tube rotates, the camera of the Cell-CT takes 500 microscopic projection images (with infinite depth of focus) over a full 360° of rotation around each cell. By analogy, a single-cell projection image is similar to a standard chest radiograph in which the entire anatomy (or cell in this case) is in superposition and in sharp focus. VisionGate's method of creating 3D images from the multitude of projections uses state-of-the-art radiologic computed tomographic reconstruction, but with visible photons rather than x-rays. The isotropic resolution of the Cell-CT reveals high-contrast, high–signal-to-noise images of cell structures in the 200-nm range. Similarly to flow cytometry, the Cell-CT analyzes 1 cell at a time. A big difference is that the Cell-CT can image in both absorption and fluorescence modes, whereas today's flow cytometry devices image by monitoring fluorescence. Nelson notes a further difference, in that “the Cell-CT measures cell morphology far more intensively and analytically than is possible with flow cytometry” and thus “can render an analysis based on a few hundred cells rather than thousands of cells,” which would perhaps be better for small numbers or microsamples of cells, such as a rare cancer cell. At 20 min per patient sample and running unattended as a fully automated system 24 hours a day, 7 days a week, the Cell-CT could analyze 25 000 samples per year. The Cell-CT translates the “radiological science of computed tomography into the cell world,” and thereby inherits decades of advancements in the science and engineering behind computed tomography. The 3D cell imaging that Nelson calls an “easy” concept was not easy in implementing, however. In radiologic computed tomography, x-rays pass through the organs being imaged with straight-line trajectories. Depending on the physical properties of the organ, the x-rays are differentially attenuated before reaching the detector. “There is insufficient density in a single cell to allow contrast using x-rays,” says Nelson. “Arguably, x-rays cannot see single cells.” The Cell-CT leverages the contrast methods available with brightfield transmission optics; however, because the wavelengths of the incident waves (in the blue to green range) are near the wavelengths of the cellular elements being detected, substantial diffraction occurs. The Cell-CT has special lenses that “bend the diffracted photons back into an image that can be seen,” explains Nelson. “In doing so, the optical image is formed by lenses that create a focal plane that is different from the projection image that is required for computed tomography.” He calls the device's ability to simulate a radiologic-projection image a “breakthrough” because it allows it to inherit all of the mathematics used in computed tomography. Another major achievement is the ability to measure orientation-independent features. Other efforts toward improving imaging of nuclear morphology have been in the development of such devices as the confocal microscope, which stacks numerous 2D images to create pseudo-3D images. These images, however, yield distorted point spread functions that constrain the features that can be measured according to their orientations. The Cell-CT generates spherical point spread functions that allow isotropic resolution. A recent report by Nelson's colleagues at Arizona State University demonstrates the use of the Cell-CT on breast epithelial cell samples (1). The authors note that this technique has the advantage of using conventional hematoxylin and eosin stains, thus making it compatible with existing histologic techniques. Dr. Eric Seibel, a research professor of mechanical engineering at the University of Washington and coinventor of the optical microscopy technology on which the Cell-CT is based points out that compatibility is very important. “Because the Cell-CT uses the exact same imaging mechanism but extends the image from 2D to 3D, pathologists and cytologists can almost immediately use this technology to make more accurate measures and earlier diagnoses,” he wrote in an e-mail message to Clinical Chemistry. Moreover, the recent report analyzed normal, benign fibrocystic, and malignant breast epithelial cells and found that unique variations in nuclear shape were associated with the malignant cells. They also found, contrary to the conventional understanding, a large degree of pleomorphism in normal cell nuclei, as well as “not-so-prominent” differences between normal and malignant cells in the nucleus-to-cytoplasm ratio. Although it is too early to predict whether larger studies will concur with their findings, the additional data from the Cell-CT could provide new parameters for the cytopathologist. Nelson makes a bolder statement: “With the combination of absorption imaging to provide diagnostic significance with fluorescent molecular markers to provide personalized disease characterization, I can see that we are about to embark on a new paradigm in microscopy that is similar to the advent of x-ray CT over 40 years ago,” he says. “That [paradigm] totally transformed the practice of radiology—I think the Cell-CT will similarly totally transform pathology.” Seibel admits that although the Cell-CT could be fully implemented into the laboratory and used “almost immediately,” it would bring about procedural change. “There are questions of how 3D images will be observed,” wrote Seibel. “Will the medical doctor and technician be aided by computers using machine vision?” Nonetheless, he sees these questions not as a roadblock, but as just a necessary maneuver to bring about change in a field of medicine that is often conservative. “[The] human pathologist will always be critical in diagnostic decision-making,” he says, “but they may not be used to hunt for the abnormal and rare cancer cells in the biopsy specimens.”