How Will AI Empower the Pathologists of the Future?
The world around us is being transformed by Artificial Intelligence (AI), whether through better cybersecurity or through the development of self-driving cars. Nearly everyone, whether they are aware of it or not, benefits from AI in their everyday lives. Medical technology is no different, presenting significant developments ranging from diagnosing strokes to more accurately reading X-ray images.
AI technology elements are experimented with in many medical fields. At this very moment, a team of experts at John Hopkins’ radiology department is working around the clock to train computers in the diagnosis of pancreatic cancer. The successful implementation of AI to spot pancreatic tumors in their earliest stages, when they are even too subtle to be noticed by most radiologists, could mean that this particularly deadly strain of cancer may become treatable.
With these hopeful and exciting prospects in mind, what impact can we expect AI to have on the pathologists of the future?
The world of pathology, which has been practicing the same diagnostic methods for the last 150 years, namely pathologists examining glass slides under a microscope, is now on the brink of bringing AI into the laboratory. The hope is for technology to increase diagnostic accuracy and therapeutic effectiveness and ultimately, save more lives. In recent years, many labs around the globe have begun digitizing tissue samples on slides. Once these slides can be viewed digitally, algorithms can be (and are already being) developed to help pathologists analyze these slides more quickly and with greater accuracy.
The Current State of the Pathology Profession
Pathologists are becoming an increasingly scarce asset. Combined with the fact that the number of cancer cases is on the rise, we face a situation where the number of cases per pathologist and the resulting workload and pressure is immense. This causes pathologists to spend less time on each case, leading to higher chances of human error, an increased number of misdiagnoses, assignment of suboptimal treatment protocols, lower patient quality of life and increase costs for the healthcare system.
Pathology services, therefore, need to adapt in order to meet the demands of the healthcare system. This means bringing cutting-edge technology into the laboratories to make the system as agile, efficient and accurate as possible.
Artificial Intelligence and Pathology
This is where AI and pathology meet.
As part of the transformation of pathology labs towards implementing state-of-the-art technologies, the practice of slide digitization, namely digital pathology, is gaining traction. Slide digitization opens up the possibility for developing AI-based pathology algorithms, using computer vision, combined with deep learning, to be able to provide the pathologists with decision-support and quality control tools, resulting in more accurate, rapid and objective diagnosis.
AI-based algorithms have the potential to support multiple steps within the pathologist’s workflow, from the earliest phase of slide quality assessment to providing a quality control step after pathologist diagnosis. AI can be used to flag scanned slides that do not meet minimal quality requirements, suggesting to re-scan them. Additionally, AI could act as a “triage” system in the laboratory, by filtering out cases without findings and by flagging those that need priority or additional attention. This allows the experts to focus on complex cases that require human evaluation.AI-based algorithms can be used to support pathologists during their review of the slides, and finally, such algorithms can provide a safety net for the pathologist, by spotting cases that have been misdiagnosed by the human eye.
The Ibex Second Read™ (SR) system is the first example of a safety-net system that is deployed in a pathology institute as part of its routine workflow. This SR system is deployed in a large institute in Israel, with the goal of detecting misdiagnosed prostate biopsies. All prostate cases are analyzed by the system and when a discrepancy with the diagnosis is detected, the case is sent back for an additional pathologist review. Indeed, the SR system has identified a suspicious PCNB that had been diagnosed earlier in the day as benign by a pathologist at the institute, allowing correction of the diagnosis before the report left the lab.
Pathologists are experts with a vast amount of training. Research has found that, in some cases, the level of human error can be reduced by as much as 85% (!) when the knowledge and experience of a human pathologist are augmented by the predictions of a deep learning system. The level and detail of information available per patient can be significantly enhanced, as computer vision and deep learning technologies are able to isolate and categorize features from a person’s tissue into patterns. In addition to supporting the pathologist in providing a diagnosis, these methods can also enhance the ability to support more granular grading and staging categories, thus enabling better prognostication of the disease, as well as potentially supporting decisions for more targeted treatment for each patient.
The Future of Pathology
With the industrial revolution of the nineteenth century, the introduction of machines left hundreds of thousands of workers redundant. Some believe that artificial intelligence will “take over” and humans will become answerable to machines, or, worse, be replaced by them. But as far as the world of pathology is concerned, this is unlikely. AI software will support pathologists during their diagnostic work, enabling them to focus their time on complex cases, while providing the tools with which to assess such cases. The algorithms will enable minimization of tedious and repetitive tasks that require minimal human input while providing data to facilitate analysis leading to accurate and rapid diagnosis.
As the shift towards digital labs continues, pathologists will find themselves working in software “labs” that are quite different from traditional laboratories. It is also likely that we will see a significant increase in pathologists’ involvement in point-of-care testing at the patient's’ bedside, leading to more involvement in making clinical decisions. And, as the analysis of tissue becomes even more precise, down to the molecular level, pathologists will have access to a wealth of data on how individual patients respond to certain treatments, allowing them to take an even more prominent role in a patient’s continued treatment and prognosis. In short, pathologists will continue in the role of diagnosticians, but with a far richer and more reliable range of technologies available to support them in a greater range of decisions.
The use of AI in pathology offers a breath of fresh air, posing great promise for wider-reaching disease prevention, more accurate predictions, and quicker diagnosis, that will result in more saved lives.
Rather than considering AI as a threat, the pathology profession must view it as an exciting and welcome addition to the lab, empowering pathologists for better, more efficient and precise decision-making.
In light of the changes AI will continue to provide, pathologists must not only excel in their own right but also keep up with the technological advancements, harnessing these new opportunities and developments to the benefit of both the profession and patients.