July 11, 2024 | For more than 150 years now, medical scientists have been hunching over microscopes turning human tissue samples into crisp colored images that help pathologists diagnose diseases. Histology, as the field is known, has made an art of the complex business of creating high-quality microscope slides—and until recently done it in precisely the same way with consistently good results.
But there are some problems that have prompted efforts in the field to “move things along,” according to Richard M. Levenson, M.D., professor in the department of pathology and laboratory medicine at the University of California, Davis. Key among these is the time it takes the most skilled technician to fix, embed, section, and stain samples so they can be examined at a microscopic level, often while lives lie in the balance.
Dozens of newfangled point-of-care (POC) histology techniques have reached such a high degree of maturity that by this time next year they could be in routine use, he says, all in the name of getting a diagnostically useful image to someone who can review it “within minutes rather than hours or days” (Modern Pathology, DOI: 10.1016/j.modpat.2024.100443). For some unfortunate patients awaiting a diagnosis, there may not be an equipped histology lab within hundreds of miles where specimens can be processed for pathological interpretation.
So many “better way” technologies are working their way into laboratory practice that they have become the subject of a special forum on point-of-care (POC) histology being held on August 19 in conjunction with the Next Generation Dx Summit in Washington, D.C. that promises to feature an abundance of attractive visuals. Levenson is co-chairing this afternoon conference with his colleague, Farzad Fereidouni, Ph.D. The pair helped develop one of the imaging systems that will be discussed and that could soon be commercially available.
Fluorescence-imitating brightfield imaging (FIBI), as it is known, is a replacement for whole-slide imaging. As with the six other approaches on the forum agenda, it introduces a novel way of getting an image of a thin slice of tissue so pathologists, or artificial intelligence (AI), can read it.
Tissue is the basis of much of today’s diagnostics, particularly in cancer, and eventually ends up as a thin slice on a glass slide, stained with hematoxylin and eosin (H&E), says Levenson. The staining process principally colors the nuclei of cells blue and other components various shades of pink.
While conventional histology is “relatively simple and straightforward,” the slow time to results and the infrastructure requirements—tissue processors, embedding centers, machines to freeze tissue so it can be cut into microscopically thin sections, microtomes (specialized precision cutting instruments) to do the slicing, and staining tools—are not insignificant drawbacks, he points out. Its equivalent is the stethoscope, which people have solely relied on for years until “magic devices” came along that use a camera to watch how the skin moves up and down with a heartbeat.
Modern-day histology practices typically produce results overnight—unless frozen sections are done, which have their own problems—because they are rather laborious, says Levenson. It all starts with a piece of tissue that gets fixed in formaldehyde and then dehydrated so paraffin wax can sub in for the water. This holds everything in place so the microtome can make sections that are 5 microns thick, roughly the size of a human red blood cell.
Since this thin section of tissue needs to be stained, the next step is to get rid of the wax using xylene and ethanol. Then, to make it a permanent slide, it gets dehydrated yet again and cover-slipped with a non-aqueous glue to preserve it for years to come.
Trailblazing Technologies
How the field moves from here to POC histology is, admittedly, tricky, says Levenson. For example, images from thick specimens need to resemble those acquired from standard 5-micron slides. Fortunately, there are now numerous alternatives to physically cutting the tissue, only capturing information from optical rather than physical slices.
One approach is reflected in the pioneering work of Rebecca R. Richards-Kortum, Ph.D., professor of bioengineering at Rice University and one of the leading lights in global health-appropriate technologies. She has described an affordable AccessPath system for immediate digital pathology of resected cervical cancer tumors in low-resource settings whereby tissue specimens get stained with inexpensive dyes and illuminated with ultraviolet excitation—a technique originally described by Fereidouni, Levenson, and others (Nature Biomedical Engineering, DOI: 10.1038/s41551-017-0165-y).
In addition to localizing fluorescence emission to a thin surface layer, the added trick in this report is the use of custom optical phase masks and associated deep learning algorithms to allow more of the irregular surface of fresh intact tissue specimens (in this case, of the vagina and cervix) to be captured in focus in a single exposure, Levenson explains. This helps get around the usual tradeoff in optics between magnification level and depth of focus, which anyone who has ever used a telephoto lens on a camera can appreciate. The microscopy with ultraviolet surface excitation (MUSE) technique she employs can be applied to any tissue specimen, he notes.
An optical imaging technique known as reflectance confocal microscopy (RCM), which uses pinholes to block out-of-focus light and thereby increase contrast and resolution from thin optical imaging planes, has been around for many years, says Levenson. As of 2016, it also has reimbursement codes that have helped bring RCM into routine use for noninvasively guiding diagnosis, treatment, and management of skin conditions including malignancies.
RCM is a no-cut option whereby a probe sits atop the skin to produce images of cellular patterns and morphology and can detect skin cancers with high sensitivity and specificity, he says. Multiple companies now offer the necessary equipment, including Thorlabs, Nikon Instruments, and Leica Microsystems.
Developments in the field took an unusual turn, according to Levenson, with the topic of light absorption, scattering, and emission (LASE) microscopy for virtual histology. He points here to the work of Roger Zemp, Ph.D., associate professor of electrical and computer engineering at the University of Alberta.
Zemp is using photoacoustic imaging in combination with autofluorescence sensing to simultaneously obtain functional and structural information, Levenson elaborates. Customarily, histology uses a probe that is fluorescently labeled so that an antibody on a piece of tissue only goes to certain places it recognizes. “You know where it went by looking for the dye [fluorescein] that was attached to the antibody.”
Here, a dye isn’t needed because “the tissue itself is fluorescent and different components have different colors,” he continues. Metabolic activity—deoxygenation indicative of tumors, for example—can be mapped based on the wavelengths at which excitation is induced. The same technique could be used to measure response to cancer therapies.
The coupling of autofluorescence with photoacoustic imaging effectively means “imaging with a sound-detecting device,” says Levenson. It involves a laser light pulsing at a frequency in the sound range, which gets absorbed by tissue components that get hotter and expand. Due to the absorbed light-induced vibrations, a sound is emitted that is detectable with a microphone.
“It’s simple because it’s stain-free,” says Levenson. “You’re making the instrument do all the hard work.”
As it turns out, important molecules (e.g., DNA, RNA, and proteins) absorb light in the ultraviolet range so the method also provides important structural information about what’s in the tissue. “It generates an image that can be converted to resemble standard H&E histology,” he says, where the nuclei in the image are blue and proteins are signaled by pink.
Levenson says that a recently developed label-free optical imaging technology known as quantitative oblique back-illumination microscopy (qOMB)—developed by his colleague, Francisco (Paco) Robles, Ph.D., associate professor of biomedical engineering at Georgia Institute of Technology and Emory University—also looks very promising. It’s a simple, fast technique Levenson describes as “an interesting mixture of high resolution but very low-cost technology” requiring no more than a few inexpensive LEDs and a regular microscope lens and monochrome camera.
The qOMB technique takes advantage of the fact that phase is a relatively easy-to-detect feature of light and can be used to image cellular-scale structure within a tissue specimen, he continues. Four different LEDs are turned on, two at a time, to produce a couple of low-contrast, grayscale images which, after applying some simple math, are transformed into a high-quality histology-scale image in monochrome that an AI tool can then convert into the familiar pink-and-blue format. “It can be handheld, and it can work ex vivo or in vivo.”
In the operating theater, it could be a great way to look at surgical margins during tumor resections, says Levenson, by either applying the probe to the surgical site while the patient is still on the table or to the surface of their freshly resected tumor.
“Fast is good,” he notes. Patients coming in for a biopsy like to get answers the same day rather than a week later, and no one wants a second operation because a tumor wasn’t entirely removed in the first place.
Currently, pathologists can check margins for cancer via a frozen section procedure. Unfortunately, “it’s hard to do, has its own instrumentation that’s needed, and is not universally available,” says Levenson. In many smaller facilities, surgeons will just cut out a tumor and “hope for the best.”
Another “big deal” POC histology technology is an open-top light-sheet microscopy platform for slide-free three-dimensional (3D) pathology of large clinical specimens, enabling whole biopsies and surgical specimens to be non-destructively imaged much like a document in a scanner, Levenson says. The approach was developed under the leadership of Jonathan T.C. Liu, Ph.D., professor of mechanical engineering at the University of Washington.
The technology was the basis of a spinout, Alpenglow Biosciences. Light-sheet microscopy is “equivalent to confocal” except a whole sheet is being illuminated at the same time versus scanning point to point; this makes it possible to more quickly acquire high-resolution 3D image data, Levenson explains.
Another colleague of his, J. Quincy Brown, Ph.D., associate professor of biomedical engineering at Tulane University, has come up with yet another way to quickly get thin section information from thick tissue. It’s called structured illumination microscopy (SIM) that involves putting a grid of light on tissue rather than a broad flashlight, says Levenson.
He adds that a confocal-caliber image is produced by moving those lines of light around and removing the out-of-focus areas computationally. At the upcoming forum, Brown will be speaking about several compelling applications of SIM in direct-to-digital on-site pathology, including large-area rapid two-dimensional imaging for core biopsy and whole-resection tumor margin assessment, among others.
The fluorescence-imitating brightfield imaging (FIBI) system being developed in Levenson’s lab works “very much like MUSE... [except] using visible light,” he continues. Thanks to “optical good fortune,” FIBI can use actual H&E dyes on thick tissue surfaces to quickly produce a high-quality, high-resolution image that looks just like standard slides—no expensive components required—and in a 100-case, 24-organ validation study had nearly identical performance (97% clinical concordance) as conventional histology.
FIBI requires “just an LED and color camera or standard microscope lens,” Levenson says. The approach has been spun out to a company called Histolix, co-founded by Levenson and Farzad Fereidouni, soon to be associate professor, department of pathology, Emory University and who will be presenting on FIBI at the upcoming forum.
AI is a “fantastic complement” to POC histology but also a “work in progress,” says Levenson. Given a digital image, AI can either help pathologists come up with a diagnosis or perform the laborious or difficult quantitative assessment tasks such as counting cells undergoing division (mitosis) or capturing location or relevant proximities of different structures or cells—for example, characterizing whether immune infiltrates are present, next to, or within foci of cancer.
AI might also take on higher-level tasks such as determining if cancer is present, its type, and whether it will respond to different drugs, he continues. “That’s the far end, where everyone is pushing this field toward, but we are not there yet.”
The current opportunities and obstacles (e.g., scarce labeled data and variations across devices and facilities) will be yet another topic of discussion at the POC histology meeting, he notes. AI can read a digital image from anywhere in the world, but “right now and for the foreseeable future, the diagnostics will be done by humans... [or with] AI and humans working together.”
Parallel efforts to update 150-year-old histology techniques need to consider the global environment where high-tech solutions can be impossible to implement, says Levenson. Recognizing that much of the world is largely “histology-free,” he and Fereidouni will be using a five-year grant from the National Cancer Institute to demonstrate the practicality of their approach in Ghana.