IVD Technology Magazine
IVDT Article Index

Originally published July, 1998

IVD Technology News

Automated visual screening systems promise faster testing—someday

Automated detection systems can easily identify biological substances tagged with radioisotopes, colored particles, or fluorescent dyes, but cytology analysis is still mostly performed by human eyes and microscopes. Cytotechnologists are trained to look for cells with irregular shapes, sizes, and patterns—tasks not easily programmed into a computer.

But such difficulties haven't prevented diagnostics manufacturers from making headway toward the acceptance of artificial visual intelligence systems for cytology screening. Pioneering the regulatory pathway is the Autopap system by Neopath, Inc. (Redmond, WA), which received FDA approval in May for use in primary screening of Pap smears for cervical cancer. The Autopap system scans slides of conventionally prepared Pap smears and ranks specimens according to their degree of abnormality. The system is designed to look for abnormalities slide by slide and to rule out the 25% of slides with the lowest risk.

A competing system in cervical cancer detection takes a slightly different approach to the screening process. The Papnet system from Neuromedical Sciences, Inc. (NSI; Suffern, NY), screens the tens of thousands of cells found on a Pap smear slide and identifies the 128 cells or cell clusters that exhibit the greatest signs of abnormality. Even in slides of normal, healthy tissues, the Papnet will distinguish the 128 cells that it registers as most abnormal. Pictures of these cells are then downloaded onto a CD-ROM. From the disk, a trained cytotechnologist can view the cells on a high-resolution video monitor.

The Neuralvision artificial intelligence software by Intelligent Medical Imaging (Palm Beach Gardens, FL) can "learn"to perform a number of common microscopy procedures. Photo Courtesy Intelligent Medical Imaging.

Although Papnet is not currently approved for primary screening in the United States, NSI executive vice president for U.S. operations, Jack Henneman, says that "clinical trials for FDA approval are under way."

NSI originally marketed the Papnet for use in a few centralized screening centers that would process samples sent in from many clinics. In response to competition, however, NSI has now modified its marketing strategy. "We are trying to get away from having clinics send their slides to a screening center," says Henneman. "We'd like each clinic to have its own Papnet."

Because of the tedium and operator error associated with manual screening of Pap smear slides, this area has been a natural first target for companies that are developing automated screening technologies. But according to experts, applications for many other diseases and infections that can be diagnosed by cellular appearance are also on the drawing boards.

The core technology of Intelligent Medical Imaging, Inc. (IMI; Palm Beach Gardens, FL), is its Neuralvision artificial intelligence software. As incorporated into the company's Micro21 microscopy workstation, the Neuralvision software examines the pixels of cell images for color and orientation, and then uses artificial intelligence to "learn" to identify specific cell types.

When the system's neural network has been "trained," the Micro21 can distinguish the major blood-cell populations stained on a microscope slide based on such visual characteristics as their color, size, texture, and density. The system learns, for example, that white blood cells (WBCs) are larger and stain darker than red blood cells, and can also determine the shape of a particular white-cell nucleus to classify it as a neutrophil or blast cell.

IMI's system displays captured images of individual cells on a high-resolution monitor for a physician to review. The Neuralvision software can perform several common microscopy procedures, including differential counts, red blood cell morphology, reticulocyte count, and white blood cell estimate. WBC differentials, the most common clinical microscopy procedure, are used to diagnose infection, anemia, leukemia, and other diseases. WBC differentials of cerebrospinal fluid aid in the diagnosis of central nervous system inflammation as occurs in meningitis.

Although the Micro21 received 510(k) clearance for WBC differentials in 1993, IMI believes the adaptability of the system's software could enable it to replace virtually all clinical microscopy procedures. "Nearly every lab procedure is now automated—with the exception of microscopy," says Sheila Thompson, IMI marketing communications specialist. "Replacing the human eye is one of the last hurdles."

NSI's Henneman thinks the future of diagnostics lies in the circuits of computer-assisted, automated systems. But when that future will arrive may depend on more than just good technology. Even with the regulatory pathway cleared, fully automated systems still face the hurdles of economic policy and adherence to tradition. Production costs for such automated systems are relatively high, making the price of automated screening greater than most insurance companies are willing to reimburse. On a per-test basis, traditional, manual screening is still generally less expensive.

Even so, says Henneman, automated screening is a technology whose time is coming. "Despite its costs, automated testing increases assay sensitivity—and that's a definite patient benefit."—Gary Woo