Digital pathology: Barriers to transitioning to the digital age

Abstract

Digital pathology (DP) entails converting physical histologic data into a digital format, ultimately enhancing data transmission, archival, and retrieval. Moreover, digitization serves as an essential prerequisite for utilizing artificial intelligence programs in histologic image analysis. Clinical use of DP has had limited uptake despite its transformative potential and numerous proposed benefits. This paper explores the commonly cited workflow, regulatory, and financial barriers currently hindering DP implementation based on the published experiences of early adopters. It provides an overview of topics including relevant regulatory bodies and guidelines, cost-to-benefit analysis, data management, and the role of artificial intelligence within DP.

Introduction

The field of pathology, which is at its essence is the study of the anatomy and functions of tissues at the microscopic level, has gone through significant evolution from its most rudimentary stage in Ancient Greece to the nineteenth century through the works of Rudolf Virchow.^1,2 The process typically entails embedding prepared tissues into paraffin wax, slicing off thin sections, mounting the sections onto glass slides, and applying chemical stains before viewing the slides for pathologic changes under a light microscope. ³ These methodologies were developed early in the twentieth century; however, they have stood the test of time. A majority of modern pathologists diagnose disease using similar methods to the pathologists a century ago. ⁴ While these techniques have persisted, the physical nature of histologic glass slides limits efficiency, collaboration, and accessibility within the field. ⁴ A traditional histology workflow entails a constant, logistically challenging cycle of physically transporting, archiving, and retrieving glass slides. Distance to a centralized laboratory has a significant impact on a pathologist's ability to receive commonly ordered ancillary testing in a timely manner, and consultation necessitates either being in a common location or shipping slides to the consulting pathologist. Contrast this with a field like radiology in which anyone with access to an EMR can view all of a patient's available studies almost instantaneously. ⁵ A radiologist (or several simultaneously) can report on images taken thousands of miles away with little delay.

Digital pathology (DP), as described by Digital Pathology Association, is the creation, viewing, management, sharing, analysis, and interpretation of digital images of glass slides and offers a novel solution to some of the innate limitations mentioned above. ⁶ This is typically accomplished using commercial slide scanners, which convert data from histologic glass slides into whole-slide images (WSIs). ⁶ Through the use of viewing software, WSIs are transmitted from a scanner to a computerized workstation and displayed on a high-definition monitor for further review. ⁷ Some of the initial benefits that come to mind when considering DP is instantaneous transmission of histologic data and the convenience of digital storage. One application that has recently experienced a surge in interest is the extraction of information from WSI and their associated metadata using artificial intelligence (AI) programs. ⁸ Figure 1 outlines a simplified digital pathology workflow and its potential applications.⁸ Like all areas of medicine, pathology also faces continual pressure to improve accuracy and efficiency. This is further exacerbated by global shortages of experienced pathologists resulting in larger, more complex caseloads. ⁹ There are high hopes that slide digitalization will revolutionize the practice of pathology by overcoming the physical limitations of traditional glass-slide workflows and in turn, address some of the most pressing concerns facing the profession.

Despite its transformative potential, DP is still considered to be in its infancy. Today, its primary use falls in research and education, and uptake into clinical practice has been slow. It is estimated that fewer than 10% of U.S. organizations have adopted DP for clinical use, and less than 5% of cases are signed out digitally. ¹⁰ The reasons for hesitation to adopt are widespread with the most cited being the high cost with an unclear return on investment (ROI).^10,11 Other commonly cited barriers include an unstable and evolving regulatory environment, disruption to existing workflows, and lack of organizational buy-in. ¹² Implementing DP demands significant resources, time, and multidisciplinary expertise beyond simply scanning slides. This review provides a concise overview of the primary barriers to clinical adoption of DP in the United States, focusing on workflow, regulatory, and financial challenges. It also delves into data management of DP workflows, and AI applications.

Current state of digital pathology

Modern DP scanners employ automated processes that reduce slide handling while enhancing scan quality and efficiency. ⁸ Unlike earlier systems, which required over 24 h to scan a single slide, the fastest contemporary scanners cite the ability to process up to 100 glass slides per hour back-to-back with minimal technical errors and human intervention. ¹³ Advances in internet speed and digital storage have kept pace with scanning technology, supporting high-volume digital workflows, and eliminating the barrier of technological capability. ⁷ The degree of digitization varies widely across pathology laboratories, forming a continuum. At one end, labs digitize cases solely for specific purposes like teaching and research, while at the other end, labs digitize the majority of their workload. Most laboratories fall into the former category, with only a small number using WSI for primary diagnosis. ¹⁴ Among these, even fewer have published about their experiences, though these reports provide valuable real-world insights into the experience of digitization.^13–21

Regulatory barriers

Despite significant regulatory progress in the United States, unsupportive regulations continue to hinder DP adoption. A primary challenge is the absence of specific legislation addressing critical aspects of a DP workflow such as WSI archival. ²² Consequently, adopters must rely on guidance from professional organizations like the College of American Pathologists and the Digital Pathology Association to navigate these current gaps. Additionally, most regulatory developments have occurred within the last decade, translating to a limited time to adapt to rapidly evolving guidelines. This section introduces the key regulatory agencies and current barriers that impede DP adoption in the United States.

Food and Drug Administration and digital pathology

The Food and Drug Administration (FDA), a part of the Department of Health and Human Services, is responsible for the regulation of medical devices. Technologies developed for DP including digital scanners, slide viewing software, and AI algorithms qualify as “medical devices” and therefore fall within FDA jurisdiction. ²³ The FDA provides oversight by issuing approvals and clearances to manufacturers that demonstrate sufficient technical and diagnostic performance. It should be noted that approval and clearance are not synonymous. Approval is generally a more rigorous process that necessitates extensive clinical trial data and documentation of safety, while clearance typically necessitates proving equivalence to an already approved device. ²⁴

Food and Drug Administration approved/cleared scanners

In 2017, the Philips IntelliSite Pathology Solution (PIPS) became the first WSI system to receive FDA clearance through the “de novo” pathway, designed for Class 3 (high risk) medical devices without existing predicates. This entailed proving that WSI displayed by the PIPS were “non-inferior” to their corresponding glass-slide counterparts when interpreted by a pathologist.^23,25 Since 2017, studies have confirmed high concordance rates (90–95%) and low discordance rates (4.5–5%) between WSIs and glass slides, supporting their diagnostic reliability. ²⁶ The PIPS is not only the scanner, but a “whole slide imaging system” which includes a high-definition display and image management system. FDA clearance is contingent on these specific components being used in conjunction with little to no room to mix and match. ⁷ Moreover, clearance is limited to in vitro diagnostic use for formalin-fixed paraffin-embedded (FFPE) specimens and specifically excludes frozen sections, cytology, and non-FFPE hematopathology specimens. Shortly after the clearance of the PIPS, scanners were reclassified from Class 3 (high risk) to Class 2 (moderate risk) medical devices. This paved the way for four additional WSI systems to gain clearance through the 510(k) pathway by demonstrating equivalence to the PIPS, a less stringent process than de novo approval, though similar restrictions on specimen types and system components apply.^27,28 Only one digital cytology system has achieved FDA clearance.^29,30

Food and Drug Administration regulatory challenges

Progress remains slow due to DP's classification as a novel technology. The FDA cites several regulatory science gaps and challenges associated with DP including “Lack of standardized test methods that allow for the correlation between technical performance and clinical performance of WSI systems” and a “Lack of statistical methods and relevant data for designing and analyzing studies of image quality in terms of pathologist or AI algorithm performance.” ²³ Consequently, there are very few vendors that have achieved FDA approval/clearance thereby limiting competition and consumer choice within the DP marketplace. ²⁸ The FDA's cautious approach prioritizes patient safety but may stifle innovation by creating a high barrier of entry for vendors.

Rapid technological advancements in WSI systems, coupled with prolonged FDA validation timelines, may deter institutions from early adoption due to concerns over premature obsolescence. For example, the Roche VENTANA DP 200 slide scanner received FDA 510(k) clearance for primary diagnostic use in June 2024. Just months later, in January 2025, the higher capacity VENTANA DP 600 (240-slide batch loading vs. the DP 200's 6-slide tray) was added to the same clearance via a special 510(k) modification. ³¹ The short interval between these clearances illustrates how quickly superior options can emerge post-approval, potentially leading labs to delay purchases in anticipation of improved hardware. As more FDA-cleared systems enter the market, greater affordability, efficiency, and interoperability are expected, further incentivizing a “wait-and-see” approach among risk-averse adopters.

An important caveat is that a WSI system does not have to be FDA cleared in order to perform primary diagnosis so long that it is validated for its intended clinical use. This being said, utilizing a non-FDA-cleared system or an FDA-cleared system outside of its intended use is considered a laboratory developed test and subject to additional regulatory requirements. ⁷ Although this is technically feasible, most laboratories are not willing to incur the additional expenses and risks of going outside the rigorous evaluation and testing of the FDA if there are reasonable alternatives.

Clinical Laboratory Improvement Amendments/Centers for Medicare and Medicaid Services and digital pathology

The Clinical Laboratory Improvement Amendments (CLIA), overseen by the Centers for Medicare and Medicaid Services (CMS) and enforced by the Joint Commission on Accreditation of Healthcare Organizations (JCAHO), establish federal standards for all clinical laboratory testing on human subjects in the US including the pathologic interpretation of digital slides. ⁶ CLIA ensures quality testing by establishing universal measures such as equipment calibration, maintenance, backups for instrument failure, and corrective actions, which apply to DP despite the absence of DP-specific regulations.^6,22 A recent regulatory development pertaining to CLIA and CMS is remote reporting of WSIs, also known as telepathology. During the COVID-19 pandemic, CMS introduced discretion, waiving the requirement for a separate CLIA certificate for each remote testing site provided that sites operated under a primary laboratory with a valid CLIA certification. ³² Post-pandemic, CMS extended this discretion on the basis that remote review of digital images is generally accurate, reliable, and timely.

Current procedural terminology codes and reimbursement

In addition to authoring CLIA, CMS plays a key role in determining compensation for laboratory services. There are 43 total current procedural terminology codes that are used to report the additional work and service requirements associated digitizing slides for primary diagnosis.^33,34 These codes are currently classified as “Category 3” which essentially means that they are “add ons” that allow for data collection associated with a service but are not associated with additional reimbursement. Reclassification to Category I “revenue generating” codes requires extensive evidence of widespread use and clinical efficacy—a process that may take years—discouraging adoption due to the lack of immediate financial incentives. ³³ Consequently, DP investment must rely on reduced operational costs and/or increased efficiency rather than increased reimbursement, highlighting a significant barrier to broader implementation. ¹⁵

Archival regulation

While traditional histology workflows mandate retaining paraffin blocks, histology slides, and cytology smears for at least 5 to 10 years per CMS guidelines, no specific regulations currently govern WSI archival. ²² Physical storage demands strict environmental controls to prevent tissue degradation, a process that is logistically challenging, time-consuming, and expensive, however, storing WSIs does not exempt labs from these physical retention requirements. This raises concern about the value of storing WSI. If images are not retained, the effort to create them may be wasted, yet maintaining both physical and digital archives introduces inefficiencies and redundancy. ⁷ Ideally, regulations would permit WSI archival alone to suffice, enabling significant cost savings for labs that digitize.

The “fully digital” misnomer

The term “fully digital” implies to a pathology workflow where the entire diagnostic process—from slide preparation to primary diagnosis—is conducted using digital tools, with no reliance on traditional microscopes for routine sign-out.^13,20 However, this term can be misleading as even advanced adopters of digitization often require glass-slide workflows for hematopathology and cytology subspecialties due to technical limitations.^7,16,21 Hematologic specimens, such as bone marrow aspirates or blood smears, demand high-resolution visualization of subtle cellular features like nuclear chromatin patterns, nucleoli, and cytoplasmic granules to differentiate cell types (blasts vs. lymphocytes). These features often require 100x oil immersion magnification, which most scanning platforms lack, thereby limiting diagnostic accuracy.^7,35 Cytologic specimens, such as fine needle aspirations, pose further challenges due to their three-dimensional nature, with dense tissue fragments, scant cellularity, and red blood cell contamination often causing scanner focusing errors. ³⁶ Z-stacking, which captures multiple focal planes to create a 3D dataset, offers diagnostic utility for such specimens but greatly increases data storage needs and scanning time. ³⁶ A large intercontinental survey on DP found that 90.2% of respondents’ scanners either did not support Z-stacking or did not use it clinically, underscoring its limited adoption. ¹⁴ Even for non-cytologic and non-hematologic specimens, digital workflows face challenges. Pathologists report difficulties visualizing subtle features including microorganisms and chromatic aberrations in WSIs, as well as the inability of most scanners to support polarized light microscopy, a feature critical for diagnosing conditions like amyloidosis.^14,20,21 An unfortunate reality is that a digital image can never exceed the quality of the original glass slide. Most pathologists consider conventional microscopy to be the gold standard of tissue diagnosis due to limitations described above and due to years, or even decades, of accumulated experience. ²⁴ It should come as no surprise that many are concerned that a switch to digital diagnosis may compromise their efficiency and/or diagnostic accuracy. Thus, many will prefer glass slides or at least require the ability to request corresponding glass slides on demand even if WSIs are available. ³⁷ Regulatory validations of FDA-cleared WSI systems further illustrate this hybrid necessity. The landmark 2017 FDA de novo approval of the PIPS relied on a multicenter pivotal study of 1992 cases, which concluded non-inferiority to glass slides with a major discordance rate of 0.6% (95% CI: −0.3% to 1.0%) across diverse anatomic sites and stains. ²⁵ However, these studies did not conclude that microscopes could be fully obviated; instead, the FDA explicitly requires labs to maintain access to conventional light microscopy for cases warranting clinical judgment. ²⁵ The most recent guidelines from the College of American Pathologists for validating WSI echo these limitations by mandating contingency plans such as rescanning or glass-slide fallback if WSI failure impacts patient care. ³⁸ Thus, while diagnostic equivalence has been established, analog backups cannot be eliminated altogether. These technical and regulatory limitations highlight the gap between the ideal of a fully digital workflow and the practical realities of current DP technology.

Financial barriers

Cost analysis is critical for determining whether clinical implementation of DP delivers institutional value, particularly in a financially constrained U.S. healthcare landscape. Many U.S. healthcare groups are operating with minimal margins due to economic inflation, administrative burdens, staffing shortages, and supply chain disruptions, exacerbated by inadequate increases in reimbursement from government payers. ³⁹ The resulting lack of capital drives organizational expenditure towards safe investments with well-established returns, often sidelining DP in the process. ³⁹ This section explores the current cost-to-benefit ratio of DP implementation.

Overall costs of digital pathology

The cost of DP implementation varies widely depending on factors such as slide volume, number of workstations for digital slide review, and image storage duration. Expenses can range from tens of thousands to multi-million-dollar investments. ⁷ Because it adds steps to the traditional histology workflow, any costs accrued from digitization are in addition to those required for traditional slide viewing. ²⁶ A detailed financial analysis at Memorial Sloan Kettering (MSK), a large tertiary cancer center producing over one million glass slides annually and an early adopter of WSI, estimated annual DP expenditure at $5,285,000. ²¹ MSK's costs included 33% for IT hardware and software ($1,744,000), 21% for scanner acquisition ($1,110,000), 19% for scan team staff ($1,004,000), 10% for IT staff ($528,500), 10% for scanner servicing ($528,500), and 7% for infrastructure staff ($370,000). ²¹ Price per slide scanned when adjusted for the cost of the scanner was estimated to be between $0.55 and $19.53 with differences in scanner “down time” accounting for the wide variance. Digital scan team staffing accounted for an additional $0.79–0.92 to the cost per slide scanned. ²¹ A 7-year cost projection for eight leading European laboratories had similar findings and reported a total expenditure of €5,087,408, with hardware and equipment accounting for 44% (€2,221,429), software 28% (€1,423,199), data storage and IT infrastructure 22% (€1,118,973), and personnel increases 6% (€323,807). ⁴⁰ The cost per slide was estimated to be 7.85€ at startup due to initial costs of equipment, but this dropped to 1.15€ by the end of the 7-year projection. ⁴⁰ Notably in both studies, costs related to the scanner itself appear to account for less than half of expenses and IT costs represent a significant portion of expenditure. A broader intercontinental survey of European and Asian laboratories found that 63% (n = 29) spent between $100,000 and $1,000,000, while 21.7% (n = 10) invested between $1,000,000 and $5,000,000 on DP, highlighting the variability in implementation costs across different settings. ¹⁴ Although DP ultimately reduces staffing requirements for physical glass slide archival and retrieval, its net effect on overall staffing remains somewhat equivocal. The transition from analog to digital workflows typically generates new staffing demands, particularly during the initial implementation phase. These include project managers to oversee the transition, IT specialists for system integration and maintenance, and technicians for slide scanning and quality assurance. ¹⁵ A large intercontinental survey highlighted marked geographic variation: 78.8% of European institutions reported hiring no additional staff, compared with only 38.5% of Asian institutions. ¹⁴ These findings underscore the substantial financial commitment required for DP, driven by both initial investments and ongoing operational expenses.

Cost savings/avoidance in digital pathology

DP offers potential cost savings and revenue generation, but its financial benefits vary across studies and remain challenging to quantify. At MSK, annual savings of $1.3 million over 5 years were attributed to a 90% reduction in glass slide requests from the departmental archive and up to a 75.4% decrease in immunohistochemistry orders. ⁴¹ Similarly, a 7-year analysis of eight European laboratories estimated a €5.29 million economic benefit against a €5.09 million investment, driven by €4.33 million from higher exam volumes, €559,000 from increased secondary consultations, €372,000 from workplace efficiency improvements, and €32,000 from reduced or avoided equipment costs. ⁴⁰ A study on the effects of digitization on task efficiency determined about 19 working hours saved per day, amounting to about 5 min in time saved per case. ⁴² Multiple studies support time savings, but this does not necessarily translate to significant reductions in turnaround time. Turnaround time is a commonly used quality metric that measures time from laboratory ascension of a specimen to the time of final report. Some laboratories report modest reductions ranging from half a day to a day, although other studies did not determine a significant difference.^18,20,40 Telepathology initiatives also demonstrate savings: UCLA's digital telepathology service saved $24 per courier trip between sites totaling $504 annually. ¹⁷ Comparably, a pathologist group in northern Ontario adopting a digital workflow reported annual savings of CA$26,000 in courier costs, CA$60,000 in travel fees, and CA$45,000 in accommodations, meals, and car rental expenses. ¹⁸

Capital equipment depreciation plays a key role in making the business case for DP adoption by spreading the high upfront costs of scanners and IT hardware over several years. The MSK study assumes linear annual scanner depreciation over 7 years and IT equipment over 5 years. When these annualized costs are divided by the number of slides scanned, the hardware cost per slide can drop as low as $0.30 if scanners are used at full capacity. ²¹ This model is unrealistic in that it assumes maximum technician efficiency and little to no scanner down-time but does demonstrate the potential feasibility of an ideal DP workflow. Despite these examples, extensive real-world data supporting a consistent positive ROI for slide digitization remains limited, as many laboratories hesitate to disclose sensitive financial details. ¹⁴

Data size and storage requirements

Storage is a critical and costly component of DP, requiring careful consideration of data volume, storage location, and retention duration. The main issue with storage in a DP workflow comes from the massive size of WSIs. The size of a WSI varies depending on the area scanned, resolution, and post-scan compression. ²¹ A “typical” WSI in a single plane with a size of 15 mm × 20 mm at 40x magnification is 15 gB (4.8 billion pixels), reduced to 1 gB with compression. ⁴³ Even with the best size-mitigating measures, a WSI is about 10 times larger than a typical radiologic image.^6,8,43 Notably, there are limitations on the extent to which an image can be compressed without affecting the image quality and/or slowing image archival and retrieval. ¹⁵ Focal plane stacking, also known as Z-stacking, adds a layer of complexity to image size. To elaborate, a typical WSI captures a singular image in a single two-dimensional focal plane, whereas Z-stacking captures multiple images at incremental focal depths within the “Z” plane. This allows for the ability of a user to scroll through images in the Z-stack, thereby mimicking the experience of adjusting focus on a traditional microscope. ⁴⁴ While this does have diagnostic utility, it comes at the cost of multiplicative data generation and greatly increased scanning times. A good way to think of this is creating a “volume” vs “surface area” of digital data. ⁴⁴

A busy laboratory may accumulate 100 s of terabytes to multiple petabytes of data in a digital workflow.^15,19,21 To put this amount of data into perspective, one petabyte is equivalent to a million gigabytes, or roughly 200,000 full length HD movies. To address storage, institutions have taken a wide variety of approaches. Some early adopters have opted to selectively store images or delete them after a period of 90 days to mitigate the high costs of long-term storage. ¹⁰ Another popular approach is the utilization of “tiered storage” in which newer images are placed in a higher tier of storage, typically onsite, which can be easily accessed. Older images are moved to lower, cheaper tiers of storage such as cloud storage or magnetic tape in which the images are still available albeit less accessible.^7,10

Other data considerations

In addition to the large size, ensuring the security of DP data adds a layer of complexity and expense. Robust security measures, including end-to-end encryption, role-based access controls, and audit trails, are essential to comply with the Health Insurance Portability and Accountability Act and safeguard protected health information.^7,15 Redundant backup systems and disaster recovery protocols further enhance resilience but can effectively double long-term storage requirements. High-resolution WSIs also demand substantial network bandwidth for efficient transmission, particularly in cloud-based or multi-site workflows.^7,15 Upload/download times can exceed several minutes per slide on standard institutional networks, potentially delaying consultations and impacting patient care. ⁴⁰ Seamless communication between scanners, storage solutions, and viewing workstations is critical to minimize latency and ensure rapid access to images during sign-out.

These technical demands underscore the need for early and substantial involvement of IT expertise, ideally during the planning phase and well before scanner selection. Institutions with constrained IT budgets or legacy infrastructure often face the greatest hurdles, as retrofitting networks, expanding storage, or implementing data security can add hundreds of thousands to initial deployment costs. Consequently, data management remains a formidable barrier to widespread DP adoption, particularly for resource-limited settings.

Whole-slide imaging formats and vendor-neutral archives

A vendor-neutral archive (VNA) is a centralized, standardized storage solution for medical images, enabling image viewing, storage, and retrieval regardless of data source or type. ¹² This is much easier said than done, as no VNA is currently capable of working with all image types or acquisition sources. ¹²

Digital Imaging and Communications in Medicine (DICOM) is a standard for communication in medical imaging and is typically associated with VNAs; however, VNAs may incorporate other non-DICOM formats. Originally developed for radiological images, DICOM has evolved to encompass imaging across all medical specialties including pathology. ⁴³ In DP, DICOM provides a standardized format for WSIs and their metadata, enhancing data archiving, retrieval, and, most critically, interoperability between systems. However, DICOM's scope extends beyond formatting, as it is purported to be dynamic, evolving standard.^43,45

Unlike radiology, where DICOM enjoys near-universal adoption, only a few DP scanners natively support DICOM while most require additional licensing or conversion software. This is largely because scanners employ proprietary tiling methods optimized for their specific hardware, resulting in vendor-specific image formats and image management systems that are often incompatible with non-native systems. ⁴⁶ Vendors have little incentive to prioritize interoperability, instead focusing on advancing their proprietary formats over vendor-neutral standards like DICOM.^12,46 The resulting fragmentation increases the risk of vendor lock-in and long-term data obsolescence—an outcome analogous to the rapid obsolescence of VHS and DVD formats in consumer media.

While comprehensive, widely supported, and successfully implemented at several institutions recently, DICOM is not without criticisms. Its main critiques center on the fact that it imbeds image data, metadata, annotations, and communication protocols into a monolithic container, therefore limiting the scalability, innovation, and interoperability that could be achieved with a more modular standard.^47,48 Furthermore, bundling all elements into one file has raised concerns about cybersecurity vulnerabilities and compatibility with modern cloud-native and modular architectures. These limitations are exacerbated by the fact that DICOM was originally designed for radiology workflows decades before the advent of cloud computing and contemporary DP requirements.^47,48

Nevertheless, DICOM remains the predominant voluntary consensus standard in DP, as evidenced by the active, multidisciplinary DICOM Working Group 26, which includes major scanner vendors, academic pathologists, and informatics experts worldwide. ⁴⁵ The present debate is therefore not whether DICOM is currently perfect for DP, but rather how to evolve it so that truly vendor-neutral systems can meet the field's complex technical and clinical demands. Ultimately, the overarching goal of vendor-neutral archiving in DP is to improve diagnostic efficiency, accuracy, collaboration, and patient safety.^43,45,49 Sustaining a unified, forward-looking approach to standards will be essential for the continued maturation and clinical adoption of DP.

Artificial intelligence in digital pathology

The use of AI in image analysis is rapidly evolving and is a major driver of digitization. Rather than immediate financial incentive, many institutions are drawn to the potential of AI in histologic analysis to optimize workflows and enhance patient diagnosis.^50–54 Digitization is a prerequisite for most AI applications, positioning labs with extensive digital workflows to lead in developing, testing, and implementing these tools. ⁸

AI excels at quantitative tasks that are tedious or time-consuming, such as measuring tumor dimensions, counting mitotic figures, and enumerating specific cell population. Since AI algorithms do not fatigue, they are able to maintain a level of speed and consistency that exceeds human capability. ⁵⁴ Another widely adopted application is computational triage. AI highlights regions likely to contain pathologic changes, enabling pathologists to focus on diagnostically challenging areas. ⁵⁴ A prominent example is FDA cleared Paige Prostate, which generates heatmaps on prostate WSIs to flag tissue suspicious for malignancy. Multiple studies have shown that pathologists using this tool achieve faster case review, higher diagnostic confidence, and equivalent or improved accuracy.^51,52 By prioritizing abnormal regions, AI effectively streamlines the workflow of pathologists, allowing them to devote more time to complex interpretive tasks. ⁵³

Despite these benefits, AI introduces important limitations and ethical challenges. One of the greatest is the threat of algorithmic bias. Because AI models learn patterns from training datasets, their performance is inherently limited by the quality, diversity, and representativeness of the data. ⁵⁴ A model trained predominantly on slides from one institution may perform poorly elsewhere due to differences in staining protocols and patient demographics. ⁵⁴ Achieving generalizability therefore requires large, multicentric, and heterogeneous training cohorts. A related issue is algorithmic transparency. Many deep-learning models function as “black boxes,” offering no insight into how individual predictions are reached. In high-stakes medical decision-making, such opacity is problematic. Algorithms should provide clear justification for their outputs so that corrective actions may be taken.^53,54

AI training presents an opportunity for labs that archive WSI to utilize, and possibly commercialize, their large data sets. While this offers a pathway to recoup DP investment, it raises serious ethical and legal questions about patient data ownership and consent. Precedents in U.S. case law have established that patients generally relinquish property rights to excised tissues and derivative data even when they are used for commercial purposes. ⁵⁴ Nevertheless, surveys consistently show that patients are far less willing to allow commercial use of their medical data than academic or clinical use. ⁵⁴ The American Medical Association's Code of Medical Ethics requires disclosure of potential commercial applications during informed consent, yet compliance remains inconsistent. ⁵⁵

A potential mitigation is comprehensive legislation such as the recently passed AI act in the European Union. The AI act classifies the AI algorithms used within DP as inherently “high risk” and therefore must abide by detailed technical, procedural, and organizational requirements. These include cybersecurity measures, human oversight, transparency, comprehensive documentation, data governance, and establishment and maintenance of a risk management system. ⁵⁶ It also necessitates that validation and training datasets are adequate to minimize the risk of bias. This is a much more robust and nuanced discussion within a rapidly developing field; however, the interested reader can refer to the cited literature for a greater picture of the uses and limitations of AI within the context of DP.^53–55

In short, AI is a major driver of digitization with great potential to augment pathologic diagnosis but should be approached with an appropriate level of skepticism and understanding of its limitations to avoid harm. As AI-driven pathology advances, robust governance frameworks for data use and patient engagement will be essential.

Status of digital pathology in Europe: A brief comparison

Although this paper primarily examines DP in the United States, Europe shares notable similarities in adoption trends and regulatory frameworks while exhibiting distinct differences in funding and implementation models. Market analyses indicate that North America holds the largest global share (∼36–46%), with Europe as the second largest (∼28–30%). ⁵⁷ Projections suggest sustained U.S. dominance through 2035, driven by private investment and rapid AI integration.

Regulatory pathways show parallels, but key shifts. Obtaining CE marking under the In Vitro Diagnostic Regulation historically allowed faster market entry than FDA clearance due to less intensive pre-market scrutiny. However, updates to medical device regulation combined with the EU AI Act have imposed stricter clinical evidence requirements, notified body involvement, and post-market surveillance.^56,58,59 These changes have extended timelines and costs, leading many vendors to prioritize FDA clearance first for its predictability and speed. ⁵⁹ The EU framework emphasizes long-term patient safety and ethical AI governance, whereas the US prioritizes innovation.

Reimbursement challenges persist in both regions, with fragmented coverage limiting scalability. ⁶⁰ Europe benefits from coordinated public funding including Europe's Beating Cancer Plan and Horizon Europe programs.^61,62 A prime example is the BIGPICTURE project, a €70 million public–private partnership funded under the Innovative Medicines Initiative, which is building the world's largest ethical repository of annotated WSIs to accelerate AI development in pathology. ⁶³ In contrast, U.S. adoption relies more heavily on industry partnerships for infrastructure and innovation. Examples include Mayo Clinic's collaboration with NVIDIA to deploy advanced computing for foundation models in DP and UCLA's longstanding partnership with Leica Biosystems for scanner integration and workflow optimization.^64,65 Large-scale federal grants for DP remain limited in the US, though there are emerging opportunities like the Rural Health Transformation Program. This allocates $50 billion over 2026–2030 for rural healthcare innovation, including telehealth and digital tools. ⁶⁶ This could enable digitization in underserved areas where subspecialty access is often lacking, and telepathology offers significant benefits. Overall, while Europe's structured funding and harmonized standards support steady progress, the US's innovation-friendly regulations and private-sector dynamics drive faster commercialization of novel technologies. The more fruitful approach is yet to be determined.

Conclusion

This paper does not aim to criticize DP or minimize its benefits but to explore why even well-resourced institutions hesitate to adopt this transformative technology. Recognizing these barriers is a critical first step towards understanding and eventually overcoming them. DP enthusiasts, adopters, and industry experts are already addressing some these barriers and developing innovative solutions. Professional organizations are actively lobbying policymakers to establish a regulatory environment that is both robust and permissive to encourage adoption. Moreover, it is reasonable to expect that DP technologies will become more affordable, accessible, and advanced in the future. Although guidance on achieving DP implementation is not extensively discussed, several comprehensive review articles address this aspect explicitly.^7,8,15,60 Some commonalities between them include a multidisciplinary approach, clear policies for data storage, and attention to quality and production control. Adopters should have a clear end goal of digitization such as integration of AI or expediting consultation.^8,60 Ultimately, early adopters will play a pivotal role in resolving current challenges, paving the way for broader implementation, and realizing DP's full potential in clinical practice.

Figure 1.

Outlines a digital pathology workflow and associated applications, simplified. Created in https://BioRender.com.⁸