Though cancer diagnosis has evolved with the latest cutting-edge innovations – from the introduction of immunohistochemistry and genomics to the use of advanced imaging techniques – further advancement is needed to support pathologists who face unprecedented challenges. Specifically, there has been a pronounced increase in both the prevalence and complexity of cancer cases, placing significant burdens on pathologists to meet patient demand. Global cancer incidence increased from 18.7 million in 2010 to 23.6 million in 2019, leading to an overwhelming influx of tissue biopsies in diagnostic laboratories (1). Moreover, as research advances and new therapies become available, diagnosis becomes increasingly complex and demands additional specialized testing.
The rise in cancer cases is compounded by a pronounced shortage of pathologists across the US and Europe. Active pathologists in the field are experiencing heavy workloads as they review more biopsies than physically possible, risking burnout and distraction. Even in pathology labs that have digitized workflows to support caseloads, prioritization, slide review, and report drafting are still handled manually and are not easily scalable. These factors increase strain on pathologists and limit their ability to provide timely and accurate diagnoses to support treatment decisions.
Enter the autopilot of pathology
The incorporation of artificial intelligence (AI) into clinical workflows across healthcare is becoming more commonplace and has proven especially useful in the field of oncology. From assisting in mammogram analysis to localizing tumors in MRI scans, cancer care has advanced significantly with the advent of AI and machine learning tools. Although the use of AI solutions in pathology workflows has lagged, deep learning technologies are particularly suited to the field and have started to have a notable impact.
AI provides rapid and accurate identification of cancer regions in slides prior to the pathologist’s review – an important benefit, but today’s solutions go even further. Tumor grading, automated measurements, classification of cancer subtypes, detection of cancer-related findings, AI-assisted reporting, and other AI-powered decision support tools increase the accuracy of diagnosis, improve the pathologist’s work experience, optimize ordering of ancillary tests, and accelerate diagnostic turnaround times.
Need a more concrete example? AI algorithms can successfully identify and distinguish between invasive ductal carcinoma, invasive lobular carcinoma, and DCIS in breast biopsies – allowing for the rapid ordering of biomarker tests, such as ER, PR, and HER2, that are necessary to complete the pathology report. AI solutions can further support pathologists in accurately scoring these biomarkers and ensuring patients receive the correct treatments. AI solutions can also provide objective Gleason grading for prostate cancer, triage cancer cases in gastric biopsies, and detect clinically relevant non-cancer features, such as H. pylori.
AI can also serve as a real-time quality control tool for pathologists. An algorithm can review slides and flag discrepancies between its findings and the pathologist’s diagnosis. In doing so, AI can help pathologists detect potential diagnostic errors; for example, missed or incorrectly graded cancers. This capability can be considered essential for those laboratories with limited personnel, high caseloads, and an inability to regularly meet auditing requirements.
AI today for a better tomorrow
The combination of expert pathologists and AI is already reshaping the practice of pathology and cancer diagnoses. By incorporating input from a variety of sub-specialties into one seamless solution, AI is becoming a real-time “digital assistant” for pathologists.
Admittedly, more work needs to be done to actualize the full potential of the digital transformation. AI solutions should be fully integrated into pathology workflows and seamlessly embedded into existing image management solutions and lab information systems. But for this to happen, the industry must make open interfaces and interoperable products its de facto standard; it must be easier and simpler for laboratories to deploy digital workflows that include best-of-breed applications from different vendors.
Imagine this lab of the future: Pathologists begin their day with cases already digitized, prioritized, and ready for review. Each case includes all details necessary for a pathologist to confirm a diagnosis, including cancer and subtype detection, tumor measurement, grading, biomarker assessment, and more. Upon reviewing the AI tool’s findings, pathologists quickly provide a diagnosis and report it to the referring physician – all using a single unified solution. With routine practices automated and more clear-cut cases completed more quickly, pathologists finally find themselves with the time to focus on more complex cases that require even more attention and expertise. And the patient? They benefit from more accurate, more comprehensive, and more timely diagnosis, while receiving more personalized care and enjoying much better outcomes.
Newsletters
Receive the latest pathology news, personalities, education, and career development – weekly to your inbox.
References
- J M Kocarnik et al., JAMA Oncol, 8, 430 (2022). PMID: 34967848
