Seeing is Believing in Image Analysis

---

By Iqbal Habib

The value of automated slide scanning and analysis systems is finally being realized. Hardware — such as automated microscopes, illumination systems and computers — is more powerful, software algorithms are more sophisticated, and network infrastructures are stronger. Meanwhile, the number of novel biomarkers under evaluation is greater than ever, creating the need for permanent digital records, and rapid information sharing and exchange. There is a need to move away from subjective qualitative data to more objective quantitative information. Together, these factors have created the perfect opportunity for growth in imaging and image-analysis technology.
This article discusses the adoption and evolution of automated slide-scanning and analysis systems, how they are being used, who is using them and how automated image analysis will shape the future of research, diagnostics and therapeutics.
In the 1990s, the Human Genome Project began generating vast amounts of quantitative data through sequencing. This data was made available to the scientific community through publicly accessible databases which in turn allowed collaboration, accelerated research and understanding of the human genome. Now, a similar explosion of data is being generated by the use of tissue microarrays (TMA). In order to foster collaboration and information sharing, databases rich with image and image-analysis data are being created within the scientific community.
Unfortunately, the explosion of new staining techniques and increasingly complex protocols has created bottlenecks for scientists trying to organize and understand this information. While humans are good at morphology, studies have shown that they are not as proficient at evaluating colour and other properties like stain intensity inherent in many of the new assays.
To more rapidly obtain and publish information, groups like The Wellcome Trust Sanger Institute (Cambridge, U.K.) have begun employing automated slide-scanning and analysis systems. In an effort to prevent “stagnation” and promote acceleration in drug research, the U.S. Food and Drug Administration (FDA) launched its Critical Path Initiative in 2004. Outlined in its report, Innovation and Stagnation: Challenge and Opportunity on the Critical Path to New Medical Products,1 the FDA described “the urgent need to modernize the medical product development process — the Critical Path — to make product development more predictable and less costly.” In subsequent events, the FDA has singled out imaging technologies as being “at the forefront” of the agency’s efforts to develop new strategies and technologies to streamline the drug approval process.2
Un-bottling the Bottleneck
Those who have not adopted automation in their laboratory can attest that manual slide analysis is a lengthy process. During the entire workflow, time is spent procuring, processing and staining tissue samples, evaluating slides through a microscope, and eventually documenting the results in a report. Much of the process can be automated by utilizing automated tissue processors and autostainers, but only up to a certain point. Beyond this point, automated slide scanning and image analysis are required to maintain a streamlined workflow.
Automated slide-analysis systems have wide research and clinical applications. In the area of research, they are being used to evaluate tissue sections stained with various immunohistochemical (IHC) markers, fluorescent in situ hybridization (FISH) probes, immunofluorescent (IF) markers and even in the general analysis of overall tissue composition. Analysis routines can be as simple as measuring the area and intensity of stain of a slide stained with hematoxylin and eosin (H & E), or as complex as evaluating a tissue microarray with four IHC markers on a single slide.
TMA analysis, in particular, is an ideal demonstration of the capability of automated slide scanning and analysis. Keeping track of each core in a 600-core TMA is complex, not to mention the need to quantitate two, three, or even four biomarkers on the same slide. Manually analyzing a TMA is taxing to a laboratory’s professional resources, and creates a large opportunity for human error: inadvertently mismatching a single core or skipping a row on a TMA is frustrating and counterproductive. Fortunately, there are imaging systems currently available that can keep track of the vast amount of data mined from TMAs.
Automated slide scanning and image analysis has many other positive implications for laboratory operations. Computer imaging of microscope slides revolutionizes how tissue samples can be viewed. In manual slide analysis, viewing a slide simultaneously with other people requires a multi-headed microscope. When using a digital image-analysis system, however, a slide can be easily scanned, digitized and shared with hundreds or even thousands of individuals via the Internet.
More importantly, automated image-analysis systems can improve the quality, objectivity and consistency of lab results. Manual slide analysis is intrinsically subjective and commonly employs broad visual scoring protocols that can yield inconsistent results. This problem has been compounded by multi-colour staining and other new complex protocols. For example, a human can evaluate the staining intensity of a cell surface marker using only four categories: 0, 1+, 2+, and 3+. In contrast, an automated imaging system can easily stratify intensity by 255 shades.
Even with the most complex protocols, automated image-analysis systems provide measurable, quantifiable information that is reliable and reproducible. Likewise, automated image-analysis systems miss fewer rare events than manual slide analysis and are less subject to the “reader fatigue,” caused by using a microscope for an extended period of time. The outcome is better, more defensible research results.
This has strong implications for the success of research grants and applications. Grant and abstract requirements increasingly call for quantitative data that cannot be obtained solely through manual slide analysis. After the scientist has identified subsets of the tissue (regions of interest) that require analysis, an image-analysis system can provide the highly detailed information necessary for a grant application.
Another powerful feature of automated image analysis is image and file sharing. As quickly as images and data are obtained, they can be disseminated via the Internet to laboratory teams and researchers across the hall, in other provinces, and around the world.
It is not uncommon for organizations to fly individuals from Tokyo to Vancouver, for example, to sit around a multi-headed microscope in a conference room to evaluate slides together. Rapid whole-slide scanning and secure networks can facilitate virtual collaboration seamlessly from anywhere in the world reducing travel time and slide-shipping costs between laboratories.
Automated slide-analysis systems are being used in a variety of settings today. University medical centers typically use such systems for translational research. Reference laboratories with high volume and short turnaround time use automation to streamline their process. Pharmaceutical companies routinely perform screenings of different treatment groups to quickly identify the beneficial or toxic effects of compounds. Clinical laboratories use the data from FDA-approved analysis systems as an aid to the pathologist. The applications are varied, the uses limitless.
Automated slide analysis will shape the future of research. Not only will it help produce results faster and easier, it will enable the extraction of image data that is unattainable by human interpretation. It will be no surprise if studies that were previously abandoned will be resurrected by employing automated image analysis.
It is time for researchers to take out the old glass slides from their archives, dust them off, scan, analyze, and instantly share the results — complete with detailed, computer-generated quantitative data — with their colleagues around the world. It may seem impossible, but it’s not. One simply has to see the images and they will quickly believe in the technology.

Iqbal Habib is the technical product manager for the pathology market at Applied Imaging Corp. (San Jose, CA), where he is responsible for marketing the company’s automated scanning and image-analysis systems to pathology and oncology markets. San Jose, Calif.-based Applied Imaging is a supplier of automated imaging and image-analysis systems for the detection and characterization of chromosomes and molecular markers in genetics and cancer applications. Applied Imaging is currently developing a system for the detection, quantification and characterization of circulating tumor cells from the blood of cancer patients. More information about can be found at www.aicorp.com.

References
1.     Medical Imaging and Drug Development. U.S. Food and Drug Administration. 6 January 2006 <http://www.fda.gov/cder/regulatory/medImaging/default.htm>.
2.     MacNeil, John S. “GenomeWeb Daily News.” Imaging Technologies Top Agenda or Critical Path Effort at FDA. 21 April 2005.