The concept of “who gets cancer treatment and what type” seems like a movie plot or a question from decades ago, but unfortunately, even in 2022, it is still a dilemma faced every day around the world. Although the underlying causes, detection and treatment options vary with the specific types of cancer, a central role of pathology is the proper diagnosis and characterization of cancer to determine the subsequent management and treatment options. The problem is that anatomical pathology, the technique used universally to diagnose cancer, is inherently subjective, with variability in both accuracy and reproducibility well documented. The promise of greater accuracy and reproducibility is driving scientists and commercial organizations to apply artificial intelligence (AI) platforms that review tens of thousands of features in the tissue and analyze every cell on the slide to ensure highly reliable results.
Standardization can save lives.
There were an estimated 18.1 million cancer cases globally in 2020, yet no way to create a completely objective and accurate determination of cancer grade, as cancer grading is dependent on the human interpretation of cells under the microscope. In short, pathology to date has been largely subjective.
For example, a patient with a slow-growing tumor could be over-called as having a more critical and advanced cancer, leading to unnecessary procedures, chemotherapy or radiation. On the other side of the spectrum, if the morphology features of a tumor are deemed to be less invasive by a pathologist, the patient could be under-called, and thus not be offered the appropriate, potentially lifesaving, treatments needed.
AI has already been adopted and is making a positive impact in radiology and patient monitoring. Recently, several organizations have emerged to combine industry knowledge and medical science expertise along with leading-edge computer science and engineering. These include PathAI, Paige, Ibex, and my own company, PreciseDx.
The issue is most critical in resource-constrained regions.
While the subjectivity of pathology is a universal problem, the issue becomes much more critical in resource-constrained regions. Countries in Africa, for example, are some of the most deprived of pathologists in the world. As of 2016, according to Dark Daily: “Mozambique has only four pathologists with a population of 25 million. Botswana has a mere three pathologists to serve its 2.1 million people.” In Mexico, there are only an estimated 1,800 pathologists for its 131 million citizens. Even the United States is not immune to the dearth of qualified pathologists, with the number of providers decreasing by nearly 18% between 2007 and 2017.
This striking data should lead us to the question: How does the limited (or complete lack of) access to pathologists impact care, and what can we do about it? By adopting and embracing AI tools, pathologists will benefit from improved efficiency and the ability to participate in directing high-quality care in other parts of the world.
Connecting patients to personalized treatment promotes the best possible outcomes. Step one in determining the best treatment is the ability to assess the risk of disease progression, metastasis or death accurately and objectively. The ability to employ AI-enabled algorithms to determine this risk, without the need to have access to local qualified pathologists, begins the process of advancing healthcare in all regions of the world, including the U.S.
In resource-constrained regions, the application of AI-enhanced pathology algorithms addresses the absence of pathologists to accurately determine which patients are more critical than others. Remote access to experts through telepathology to review and analyze the pathology slides, in conjunction with AI-enabled risk assessments, could provide the needed information to make better treatment decisions. Without such access to experts, local providers are often forced to make unsupported decisions about who will have access to the limited treatment options they can offer.
These providers deserve access to the support tools necessary to optimize the distribution of care across their patient population.
Technology can open access to care worldwide.
Through technology—specifically the digitization of slides and the use of AI—we can make pathology and pathology insights available to nearly everyone. AI has taken off in many aspects of healthcare, and in the past five years, there have been several new technologies introduced that are dedicated to assisting pathologists and oncologists. These new technologies include several oncology solutions from Philips, histology solutions from Leica Biosystems, and detection and monitoring solutions from Hamamatsu Photonics.
While still in its early stages, AI can be trained to “data mine” millions of data points to identify and quantify key morphologic features and cellular characteristics for each cancer type. These techniques are being applied using whole slide imaging (WSI) so every cell on the slide is looked at, not just a sampling from the slide. This creates a much more statistically robust set of data, and the AI can present the results in both absolute and percentile format. Access to this data and the information-rich analysis that can result has the potential to support pathology and oncology in new and exciting ways by providing highly accurate, objective patient-specific guidance.
The potential is for improved care for everyone. This is not an instant fix; however, resource-constrained regions can best take advantage of this new technology by putting in place the required tools and training for technicians to process the tissue for the scanners.
Healthcare industry members should pay close attention to the publications coming out, and, along with their colleagues, become early adopters of those processes in which artificial intelligence will allow improvement in outcomes and increase efficiency.
To fully take advantage of this process, the industry will need to adopt infrastructure with the bandwidth to send large images and image-based, annotated results to fully leverage and integrate the power of artificial intelligence in healthcare.
With technology, geographic location becomes significantly less important in relation to the quality of care. An internet connection can open access for everyone to new levels of accuracy and standardization in pathology. As just one of the many opportunities for AI in healthcare, the inclusion of AI in cancer pathology has tremendous potential to support highly personalized treatment and better patient outcomes.