Digital and Computational Pathology Are Pathologists’ Physician Extenders

“Physician extender” is a long-used term in medicine¹  that denotes a “non-medical healthcare provider who sees patients on behalf of or in conjunction with a lead physician or physician(s).”²  Increasing physician demand requires continuing assessment of the value of physician extenders, including within pathology.^3,4  During the COVID-19 pandemic, health care institutions and private practice groups across all medical fields felt the unprecedented demand for fast and accurate medical laboratory testing and diagnostics.^5–7  Beyond the COVID-19 pandemic, the urgency for pathologists and medical laboratories across the world to continue to deliver fast and accurate diagnoses and laboratory results has persisted, and is constantly being reinforced as staffing shortages continue to grow.^8,9 

Although demand for pathologists’ services has remained constant or increased, the growing pathologist shortage has placed substantial stress on all pathologists, including new-in-practice pathologists.^10,11  Pathologists are faced with increasing demands and decreasing workforce supply, resulting in pathologists and trainees feeling the excess “weight” of these demands, for example, increased productivity expectations without necessarily increased compensation.¹²  Without pathologists’ exceptional, patient-centered care via timely anatomic and clinical pathology reports, health care institutions will fail and patients will suffer tremendously.¹³  This was clearly recognized during the COVID-19 pandemic and in previous public health emergencies.^14–18  Despite institutions’ current financial constraints, more investment must be directed toward building and supporting pathology departments and laboratories. If this is not done, anatomic and clinical pathology diagnostic reports will be further delayed, leading to increasing levels of physician and patient dissatisfaction. Indeed, if health care institutions desire to improve patient-centered care and the professional lives of their physicians and staff, substantial efforts must be directed toward ensuring exceptional pathology and laboratory medicine services.

The pathologist shortage is not readily remediable and is likely to worsen. The journey to becoming a pathologist is long. Pathologists, like other doctors, go to medical school (often 4–6 years depending on location), and following medical school they complete residency training and fellowship (an additional 4–6 years). Many pathologists who practice in the United States have practiced previously in other countries and go through rigorous financial lengths to obtain US medical licensure, residency training, and legal citizenship to practice in the United States. Although recent pipeline efforts have focused on increasing US medical student and international medical graduate exposure to pathology (because many medical schools lack adequate exposure within their curricula),^19–22  it will be many years before health care institutions bear the fruit of these efforts. As such, it behooves institutions and pathologist groups to recognize the scarcity of pathologists and their increasing level of interchangeability because today’s molecular-based medicine increasingly demands subspecialty pathologists for diagnostic accuracy. Furthermore, pathologists need to be supported financially and technologically to best ensure their ability to provide diagnoses in an accurate and timely manner, and to best ensure such for our future.

Digital and computational pathology through whole slide imaging (WSI) and integration of artificial intelligence (AI) has been clinically shown to be equivalent in allowing pathologists to make primary diagnoses on computer screens in similar amounts of time and just as accurately compared with traditional glass slides on physical microscopes.^23–26  The benefits of digital and computational pathology cannot be underestimated because they are likely to serve as extenders in 3 major ways.^27–30 

First, digital and computational pathology enables flexible pathology practice and promotes a workforce that is not restricted by time or location (allowing for remote work). Digital slides on computer screens can be manipulated seemingly with clicks and annotations (eg, on a Microsoft PowerPoint or PDF), regardless of location, and can be emailed in seconds for consultation or academic venues (ie, posters, conferences). Advancements in costs, resolution, and storage space strategies continue to change the direction of digital slide utility in diagnostic practice and education, especially among tech-savvy new-in-practice pathologists.³¹  Evidence-based practice guidelines for validating WSI imaging systems have been established, and pathology departments should follow these when instituting WSI.³²  This flexibility also helps to mitigate potential hospital pathologist office space capacity limit concerns while improving return on investment and overall health care finances.³³  Glass slides may get lost or broken in transit or misplaced. Increasingly affordable WSI slide scanners and AI-diagnostic technologies produce non–geographically confined pathologists, so that fewer hospital pathologist offices would be necessary,^34,35  and the freed-up space could be used for other essential patient care needs.

Additionally, appropriate implementation of pathology AI-driven practice models will result in long-term practice efficiency that will help offset rising caseloads and pathologist workforce shortages, and potentially decrease pathologists’ burnout risk by reducing time spent on cumbersome paperwork (ie, pathologists may spend more time “being pathologists” rather than navigating administrative tasks).³⁶  Integrating more practical, innovative, and efficient means of excellent patient care, while reconstructing the value of pathology practice, may help prevent health care systems from losing millions of dollars annually due to workforce turnover or shortages.^37,38  The estimated digital pathology market revenue is expected to be $1.8 billion by 2028.³⁹  Pathology and laboratory medicine contribute significant revenue to health care systems,⁴⁰  and implementing widespread digital and computational pathology services is vital to improving financial savings while optimizing processing logistics and minimizing pathologist labor.³⁴ 

Second, AI-enhanced computational pathology will help alleviate rote tasks that often require significant pathologist time, such as question-and-answer type review,⁴¹  mitotic figure counting,⁴²  accelerating chart review,⁴³  improving immunostain recommendations,⁴⁴  and triaging cases.⁴⁵  AI and digital pathology can mitigate the potential for human error, although AI-based diagnostic tools may exhibit other errors and patient care data safety concerns that would require addressing.⁴⁶  AI-driven decision-making via machine learning occurs using deep-learning, neural network–based algorithms (“black box theory”) that cannot be entirely described by human beings or AI technologies themselves.⁴⁷  Medicolegal concerns may arise (eg, pathologist liability with delegating and describing AI-guided decisions in court), but initial proof-of-concept AI studies (particularly in cancer evaluation) and continued emphasis on randomized controlled trials will help to establish standard ethical guidelines that may be widely adopted.⁴⁷ 

Finally, AI diagnostic tools may be trained to significantly improve diagnostic accuracy, pathology report turnaround times, and overall clinical decision-making in pathologist-specific diagnostic practice.⁴⁸  Pathologist buy-in to the added utility of AI diagnostic tools is important, and a recent study by Meyer et al⁴⁹  suggests that many pathologists are willing to rely on AI data–driven input to aid in clinical decision-making (eg, detecting cancer on prostate core biopsies). Because of accelerated advances during the COVID-19 pandemic, digital pathology and AI may be able to safeguard clinical services, pathology-based research, and education; however, challenges to widespread adoption persist, such as public perception of data sharing, standard imaging formats, technology dependability, and costs of WSI implementation.⁵⁰  These challenges will undoubtedly be addressed.

Berbís et al⁵¹  predict that by 2030 digital and computational pathology systems will be routinely used within pathologists’ diagnostic workflows. Computers will not replace pathologists, and pathologists must be prepared to evolve their practice behavior by adapting to the integration of modern technology.⁵²  Pathologists are facing continuously increasing diagnostic volumes amid a continually decreasing pathologist workforce. An almost 18% decrease in pathologists from 2007 to 2017 highlights this.⁵³  There are workforce shortages of other medical laboratory professionals required for efficient diagnostic workflow and excellent patient care, such as pathologists’ assistants (PAs), histotechnologists, and cytotechnologists.^54–56  Digital and computational pathology is a catalyst in addressing those shortages effectively. Training now and future generations of pathologists in digital and computational pathology environments is critical for growth. Failing to adopt these technologies now will contribute to exponential deficiencies in present and future health care delivery as demand for pathology and laboratory medicine services continues to outpace workforce supply.

Digital and computational pathology—as with any emergent technology—has its own vital concerns that must be addressed.⁵⁷  These include but are not limited to: (1) the cost of high-throughput digital slide scanners that can scan thousands of slides per day, (2) computer hardware storage space (ie, 1 glass slide can use more than 1 GB of space), (3) employee training programs to establish new workflows (eg, someone must scan the slides after they are stained and fixed, and the pathologist must know how to access/scan the slide from their computer screen), and (4) medicolegal/ethical concerns toward routine, validated, and acceptable digital and computational pathology workflow practices.^27–29  Although these concerns exist, research is increasingly focused on the effective, safe, and just integration of digital and computational pathology workflow services into routine diagnostic pathology practice.^30,58–60  Much of this focus regards regulating computational pathology in routine diagnostic applications, namely via sufficient training of machine learning (ML) models with high-quality (and even low-quality) tissue to ensure accuracy and consistent avoidance of inaccurate diagnostic decisions.⁶¹ 

Additional focus regards the harmonious standardization of digital pathology technologies, such as by the Alliance for Digital Pathology (or the “Alliance”), to ensure precise, synergistic, science-driven data to establish regulatory guidance and improve clinical-decision making for stakeholders and their patients.⁶²  Redrup Hill et al⁶³  have identified 6 general medicolegal regulatory ethical concerns that may help stakeholders guide the use of AI-driven digital pathology tools as well: (1) risk and potential harms; (2) impact on human experts; (3) equity and bias; (4) transparency and oversight; (5) patient information and choice; and (6) moral ownership of error. We must also consider the increased regulatory burden that implementing AI and ML-enabled medical devices, including WSI-related devices, may place on countries as AI in health care continues to be addressed by governmental agencies.^64,65  According to the US Food and Drug Administration (FDA), these models and their diagnostic algorithms fall under the classification of “software as medical devices” (SaMD),⁶⁶  which is software used for medical purposes without being a part of a medical devices’ hardware, as defined by the International Medical Device Regulators Forum (IMDRF).⁶⁷  SaMDs require significant regulatory oversight from agencies like the FDA before they can be routinely and ethically implemented in health care institutions under policy-driven guidance. Medical device regulatory groups like the IMDRF develop internationally agreed-upon documents and provide frameworks to institute the guidance of SaMDs at a global level.^66,67 

In 2023, the FDA updated its publicly available list of approved AI/ML-enabled medical devices that meet appropriate, regulated marketing standards in the United States, which includes 4 pathology-related products (“Paige Prostate” by Paige.AI; “Tissue of Origin Test Kit FFPE” and “Pathwork Tissue or Origin Test” by Pathwork Diagnostics Inc; and “PAPNET Testing System” by Neuromedical Systems Inc).⁶⁸  Time to FDA approval for any device may take anywhere from weeks to years,⁶⁹  but recently proposed regulatory frameworks by the FDA hope to refine this approval process for AI/ML-related medical devices as their influence in health care institutions, including pathology and laboratory medicine departments, continues to rise.⁷⁰  Although there is great potential for digital and computational pathology models to continually serve as pathologists’ extenders, major barriers to widespread adoption currently exist (such as ethical regulatory guidance and time to marketing clearance) that will hinder everyone from quickly adopting these tools, even if they wanted to.⁷¹  This includes the development and use of internally created AI/ML medical devices and potential delays in regulatory clearance (up to years) of vendor SaMD models for any clinical lab. Above all, our primary role as physicians is patient safety, and routine AI/ML-driven digital pathology services cannot be justifiably and routinely used unless there are standard guidelines to ensure safety, which are continually being addressed by stakeholders. Evidential support is accumulating to justify the investment required of digital and computational pathology so pathologists can continue to deliver timely and exceptional patient-centered care in our resource-limited time.⁷² 

PAs, cytotechnologists, histotechnologists, and other medical laboratory professionals have been suggested as substitutes for pathologists to render a final diagnosis.^73–75  That would cause significant diagnostic confusion throughout the health care system.⁷⁶  Nonphysician health care professionals who “have been shown to have the skills necessary” may, in theory, be equipped with additional training to take on, with physician oversight, tasks routinely vetted through physicians.^77,78  However, the duties and roles of PAs, cytotechnologists, histotechnologists, and other medical laboratory professionals are proportional to their specific educational training, as with other nonphysician health care professionals.^79,80  Their professional roles are part of the continuum of patient care, just as pathology and laboratory medicine are more broadly a part of this continuum.

Their educational training, however, does not allow them to step into the diagnostic role that is the role of the pathologist. Pathologists can render accurate, proper, and timely patient-centered diagnoses. To do so requires fundamental medical and surgical organ-based clinicopathologic understanding vetted through multiple years of rigorous medical school and residency/fellowship training, with exposure to pertinent subspecialties during preclerkship and clerkship training and successful passing of national board exams at all levels. Physicians not trained as pathologists are not able to practice pathology⁸¹ ; other nonpathologist laboratory professionals are not qualified to perform the role of the pathologist either. This role of the pathologist, in the continuum of patient care, simply cannot be duplicated by another health care professional.

In conclusion, we are in a time of increased demand for timely and accurate anatomic and clinical diagnoses by pathologists. With continually increasing tissue specimen and clinical laboratory volumes and a shrinking pathologist workforce, pathologists are stretched thin.⁸²  Actionable measures must be taken. Routine digital and computational pathology services must be instituted to assist pathologists in making accurate and timely diagnoses. These extender tools can truly serve pathologists in providing timely and patient-centered care through flexible sign-out capabilities, assistance with morphology assessment, and the potential to help automate routine workflow in other academic and professional areas. Concerns of time, costs, and medicolegal ethics are being addressed effectively,⁸³  and studies continue to validate routine integration of digital and computational pathology services.

Health care institutions must now focus on enhancing their pathology departments for successful and efficient futures. Otherwise, health care institutions will continue to experience lengthy diagnostic turnaround times, which in turn will result in significant health care professional and patient dissatisfaction and decreased internal revenue. Ultimately, it is reasonable to envision an institution that does not institute digital and computational pathology, or no pathologists retained, with patients receiving suboptimal, inaccurate, delayed, or even no diagnoses. We must be optimistic regarding the institutionalization of digital pathology in routine diagnostic pathology practice because we are currently living in such an era that will only continue to emphasize the implementation of AI/ML medical devices and other computational pathology services in health care institutions and training programs.⁸⁴  Nevertheless, barriers to widespread adoption still exist, but pathologists may be able to continue mitigating these barriers through increased advocacy efforts at institutional, state, and national levels. Now is the time to improve pathologists’ professional lives with the additional digital and computational tools needed to appropriately address increasing health care demands and provide exceptional patient-centered care.