Digital pathology began in its earliest form as digital microscopy. These devices were composed of cameras integrated with a conventional microscope to capture images through the objective lenses.1 Digital microscopes allowed for both static image capture or live image transmission to an external monitor or output device. These images would represent a small fraction of the tissue present on the glass slide being limited to only 1 field of view (FOV). The first whole slide scanners (WSS), were quite primitive compared with their contemporary counterparts. Early efforts to introduce whole slide imaging started in the late 1990s, where Ferreira at the University of Maryland Institute for Advanced Computer Studies and Joel Saltz from John Hopkins University designed the first software system to support whole slide images (WSI).2 This system used a robotic microscope and computer software to stitch individual static images of a slide, captured in a tile-based mosaic pattern. Its limitations included a lengthy time to acquire all the images, and each focal plane would generate 7 GB of data. Through advancements in automation, scanner throughput, superior lenses and objectives, robotic engineering, improved camera sensors with high-resolution capabilities, and larger digital storage capacities, WSS were developed to generate digitized images of routine histopathologic lab-generated glass slides.3 Telepathology advancements and robotic microscopy permitted the pathologist to remotely view and control a glass slide as well as WSI. Numerous vendors now are mass-producing WSS with different features and capabilities dependent on laboratory needs. WSS are able to scan slides in minutes, with high throughput capacities, and manage WSI in a dynamic manner. Worldwide, there are over 30 companies providing digital pathology hardware, software, and services, with a global market expected to reach ∼$4.5 billion by 2018.4
The digital pathology ecosystem consists of 3 major components: information systems, digital pathology system (DPS), and system tools (Fig. 1). Ideally these systems would support interoperability in an integrated digital workflow between the hospital information systems (HIS), electronic medical record (EMR), laboratory information system (LIS), and Picture Archiving and Communication System (PACS). The DPS plays a central role in this ecosystem whereby the pixel data flow from an acquisition device (ie, hardware) to a viewer application (ie, software). Enabling a digital pathology workflow. System tools further integrate the pathologist’s workflow to analyze WSI such as image analysis or computer-assisted diagnosis.
The DPS process follows ARMS: acquisition, retrieval/storage, manipulation, and sharing of the captured pixel information from the glass slides (Fig. 2). This cycle is the mainstay of processing and managing the digital images captured from WSS.
The hardware supports the image acquisition process in capturing the picture elements. Camera sensors capture the color information from the pixel data. Files are generated from the pixel data (ie, WSI) as well as associated metadata captured from the scanner slide label macrocamera (eg, patient health information, barcode decoding). Retrieval and storage of the WSI files can be managed dependent on the scale of the intended use case (ie, local storage or servers). Manipulation of the WSI refers to their usability for various applications. WSI can be shared for internal use (eg, multidisciplinary tumor board) or externally (eg, consultation). Increased availability of sharing will increase communication, value added reporting, and documentation to improve patient care.
The growing digital pathology adoption and diversity of use cases has enabled digital pathology vendors to develop and support several whole slide scanning hardware. The digitization of glass slides can be performed through digital cameras integrated with conventional microscopes, WSS, and dynamic-robotic WSI devices. Slide input bays vary as different vendors offer varying inputs; from tray-like loading bays, to proprietary cartridges, whereas others have a higher level of interoperability and can use laboratory stainer slide racks as the input bay of the WSS. The slide feeder presents the slide for image acquisition, which requires essentially the same components found in a conventional microscope (ie, light source, imaging optics, and stage) along with digital imaging sensors to convert the light to a digital signal and software for image processing and composition. The light source may be brightfield (ie, light emitting diode, bulb) or may even utilize immunofluorescence channels/filters which allow for quantification of single or multiplexed stained proteins. The optics may contain 1 or several lens objectives, ranging in magnification from ×1.25 to ×100 oil. The glass slide is placed on a stage, covering the x and y-direction during the scanning process. WSS predominantly exist in 2 forms: line-based or tile-based scanners. Line-based scanners utilize a slide stage that moves in a linear path across 1 axis of the glass slides for acquisition. Sequential and complete coverage of the glass slide allows each image strip to be assembled as 1 large image file. Line scanning has become a preferred method of slide acquisition due to the simplification of the image alignment process; the degrees of freedom associated with each individual image are greatly lower than tile-based scanning. Also, using a motorized glass slide stage, tile-based scanning obtains square image frames that are programmatically stitched together, with proper alignment, into a single image as a mosaic pattern. Although tile-based scanning systems are much more complex than the line-based systems, current WSS have enhanced engineering, computing technology, lighting, and real-time quality assessment to allow for efficient and accurate image acquisition. Alternative image acquisition methods to decouple the image capture and image focusing for ultrarapid processing, such as array microscopy scanning, have been explored.5
Each imaging device provides a distinct use case, such that scanner specifications, throughput, services, and cost play a central role in assessment. Five main categories of digital pathology hardware currently exist in the vendor market: integrated microscopes, low throughput (less than tens of slides), high throughput (hundreds of slides), real-time dynamic-robotic scanners, and nonbrightfield scanners (eg, fluorescence).
Integrated microscopy has been the gateway into digital pathology. Integrated microscopes are digital cameras connected via a c-mount to a traditional microscope. These systems initially started as capturing static images from the objective lens of the microscope, however were limited to a single FOV per image. Images can be captured through desktop software, which then are stored and shareable through mechanisms as simple as email and also through dedicated telepathology Web sites. With smartphones or other mobile devices, microscope attachment adapters facilitate placement of the device’s camera to capture microscopy through the microscope’s eyepiece, whereby images are shareable through electronic formats (SMS, email, social media).6 Cameras for integrated microscopes can be either live streaming or dedicated for static capture.
Software is currently available to couple integrated cameras and image stitching software algorithms to allow for real-time capturing of user-defined FOVs. As the operator reviews the glass slide on their conventional brightfield microscope, the images are being instantaneously stitched together into a single image. This methodology is operator dependent, as the image properties (ie, FOV, magnification, blur) are dependent sequelae. However, this offers flexibility in multiple z-plane and FOV digitization, and generates smaller file sizes. The generated files can be interoperated with other existing platforms and software available for image management, annotation, and sharing. Integrated microscopes offer the lowest throughput, however also incurs the least cost.
Low Throughput Whole Slide Scanners
Low throughput devices are generally used for low to medium volume scanning requirements. Other uses of low throughput scanners are for special niche service needs that high throughput scanners otherwise do not readily deliver (ie, larger glass slide formats, dedicated high-resolution imaging). The benefit of lower throughput scanners is the advantage in exhausting a smaller footprint of the laboratory space. The mechanics and engineering are similar to their higher throughput counterparts; however cost is generally lower than the typical higher throughput WSS.
High Throughput Whole Slide Scanners
WSS may be designed to scan a relative few number of slides, while other larger hardware can accommodate hundreds of glass slides. High throughput scanners are typically large volume core laboratories that aspire to digitally archive or digitize prospective clinical cases. Both manual and automated slide input bays are available. These scanners typically have slide input capacities between 100 and 1000 slides, with some that offer continuous loading capabilities. True continuous loading offers a “load and walk away” workflow, and where the scanning does not necessitate interruption while loading new glass slides. High throughput scanners are able to work with conventional slides, 1×3 inches (25×75 mm)×1-mm thick. Some WSS offer larger, 2×3 inch slide scanning or up to 8×6 inch (200×150 mm). The slide label can be captured via a separate macrocamera, which also deciphers 1D and 2D barcodes. Most WSS in this category offer image resolution at 20× equivalent magnification ranging from 0.24 to 0.5 μm per pixel and ×40 equivalent magnification: 0.1375 to 0.25 μm/pixel (lower microns/pixel equate to higher resolution). Slide scanning speeds are dependent on the engineering and robotics of the hardware. Total scanning time factors include, but are not limited to: the imaging acquisition process, sensor size, tissue scan area, objective lens/equivalent magnification, automation, motorized stage and robotics speed, focusing method, number of z-planes to image, and network connectivity.
Dynamic-Robotic Imaging Devices
Real-time dynamic-robotic imaging devices are increasingly being used for telepathology (ie, intraoperative diagnosis). These scanners have low throughput (1 to 5 slides per bay), as generally each frozen section case does not generate many slides. These scanners have dedicated computer connections with vendor software to view and control the robotic microscope in real time. Pathologists can connect to the computer over a secure network and operate the microscope viewing the glass slide in real time (ie, pan, zoom, focus). This enables pathologists to provide remote evaluations, such as during intraoperative consultation and rapid on-site cytologic evaluation7 while they are at another location. The objective lenses and resolution of these scanners are similar to the conventional microscope, offering a range of objectives to view the glass slide with. Advances in software allow for multiple and different FOVs of the same or different slides to be viewed simultaneously. Z-plane viewing can be more effective while live-viewing the glass slide, since the pathologist operating the scanner is able to focus different z-planes efficiently in real time. These WSS do have full scanning functionality, although lacking high capacity slide bays, are scanned when necessitated.
Navigating WSI comparable to a glass slide on a microscope is crucial to the DPS and enables their manipulation in other downstream systems. The software to navigate the WSI is referred to as the WSI viewer. Users can view and navigate (ie, pan, magnify, focus) WSI on a digital monitor.8 Basic navigation features allow the pathologist to pan around the image, usually providing a smaller thumbnail overview image for the user to observe where on the WSI they are viewing. Also, standard magnification scaling of the WSI is available to view the digital slide at higher resolution/magnification. Some viewers offer z-plane (z-stack) viewing, where a virtual slider is presented to change z-planes if the slide was scanned in such format. Furthermore, WSI viewers allow possibilities that conventional brightfield microscopy does not. Additional features include viewing WSI at unconventional magnifications (ie, ×1, ×15), screenshot functionality, virtual annotation tools (eg, bookmarked regions of interest, measurements, overlay text/objects/shapes), or more sophisticated features such as coregistration (eg, synchronization of multiple WSI), heatmap-like slide coverage, and image analysis. Teleconferencing tools are available for simultaneous digital slide viewing of multiple remote users. Some viewer software allows for multiple active displays to view either a different FOV of the same slide, or a completely different slide. Coregistration, enables overlaying WSI atop each other, this allows immunohistochemical stains of interest to be superimposed on the original digitized hematoxylin and eosin (H&E) slide, which may aid in precise identification of cellular protein expression. Additional image analysis tools are evolving, including quantitative and qualitative measurements. Integration of computer-assisted diagnostic algorithms into viewer software provides the intersection between digital pathology and the new field of computational pathology (ie, machine learning, artificial intelligence). Viewer software vary in interoperability, specific WSI viewers are required to view certain proprietary WSI formats, however web-based and other viewer applications use open-source tools to allow for vendor-agnostic viewing of WSI file formats in a single viewer.9,10 Some viewers are compatible with mobile device use (ie, smartphone, tablet). Integration with LIS is currently limited to few systems, however as DPSs mature, increased integration will surely follow.
Compared with the radiology counterparts, an average standard pathology WSI file size can range from a few hundred megabytes to several gigabytes. WSIs are formatted as multiresolution pyramid composed of hundreds of thousands of individual images optimized for rapid rendering and real-time viewing. This encodes a multilayered pyramid model where either the specific FOV is held constant, and the subsequent images are derived from adjacent image data. A list of vendors, scanners, and file formats are shown in Table 1. Several vendors utilize proprietary image file extensions, however there are ongoing interoperability initiatives.9–11 Various compression formats are used to generate smaller file sizes. Image files can use various compression schema such as TIFF which can be lossless or lossy compression, or JPEG which uses a lossy compression format. Digital Imaging and Communications in Medicine (DICOM) has published Supplements 145 and 122 to discuss a universal WSI interchange standard, and is an ongoing effort. Dedicated WSI viewers are needed to view proprietary file formats, however progress has been made to cross-view different WSI file formats on single viewers.
As vendors develop proprietary WSI file formats and software, features of each viewer may be limited to their own vendor format. Differing WSI file formats curtail interoperability if their respective WSI file format is not supported by another vendor viewer software. Viewer software can occur as proprietary stand-alone desktop applications or web-based; some of which are able to open multiple WSI file formats. Open-source software (eg, OpenSlide, Open Microscopy Environment) have been developed and allow for compatibility with many WSI file formats. In contrast to stand-alone desktop applications and web-based platforms, some vendors have developed image management systems to support higher level clinical integration.
Another focus of interoperability stems from the DICOM standard, which includes a core file format and networking protocol. Currently used predominantly in radiology, these systems support DICOM files that include the medical image of interest as well as other patient data such as name, sex, date of birth, clinical history, the imaging device used and settings. The DICOM network protocol is used by imaging devices to exchange data. The images and patient information can safely and securely be transferred over a network. Imaging devices communicate over the network in a client-server architecture and can query, archive, or import supported data into the PACS or image management platforms.
The PACS allows viewing of images as well as multiple users at multiple sites to view the images concurrently. This connectivity permits workflow integration including scheduling procedures and downstream report tracking. Hospital medical imaging systems may have infrastructure to store and manage digital medical images in a PACS or vendor neutral archive that serve individual images to specified applications. File storage can include radiology images, pathology WSI, gross pathology, dermatology, ophthalmology, cardiology, endoscopy images, and surgical image files. Notable success of DICOM is rooted in the ability of imaging devices of multiple vendors to integrate and interface together based on a single standard. Initiatives to support DICOM interoperability in digital pathology have been initiated and are underway. DICOM Working Group 26 is 1 of the 32 Working groups established by the DICOM Standards Committee, and explicitly deals with the field of Pathology. Through the Working Group, DICOM supplements 145 and 122 provide a framework to incorporate pathology specimen identification and accessioning, and WSI storage and retrieval. A DICOM standard could allow for pathology designed PACS-like digital workflow to increase efficiency and communication. Current WSI file formats are larger than their radiology counterparts, however research investigating potential image compression techniques may allow for efficient and cost-effective data storage for WSI without loss of pixel data required for pathologic diagnosis.12
Current standard methodology in navigating WSI viewers requires the pathologist to use a conventional computer mouse to either click-hold-pan each FOV of the WSI. Software enhancements have allowed for using the mouse as the pan function itself (moving the mouse without clicking will move that direction in the viewer) or being able to click and drag on the thumbnail overview image for more efficient navigation. Other input devices have been tested with WSI viewers to establish the most ergonomic and rapid viewing. Users may utilize a standard computer mouse, trackpad, trackball, joystick, 6 df navigator, or any other input device that communicates with the viewer; some input devices may require device mapping for improved functionality. One study evaluating several input devices identified the 6 df mouse in having the least time spent per slide, although this finding was not statistically significant.13 Input devices for viewing WSI will be an integral part of pathologist’s daily workflow.
Applications for digital pathology are highlighted in Table 2. Continued innovation to further develop digital pathology applications are driving adoption and implementation of innovative systems. Traditionally, glass slides have been the mainstay in conventional pathology workflow. In modern practice with increasing clinical volume, turnaround time pressure, and value-based care; glass slides are impractical to archive, manage, share, and easily scale for use. Digital pathology can allow for efficient diagnosis and management of cases as well as facilitate education for all clinical and patient use (ie, digital tumor boards).14 Research applications could open the field to medical advances in prognostication of disease, quality improvements, and computer-assisted diagnosis through machine learning or artificial intelligence algorithms developed through computational pathology. In a fully implemented DPS, interoperability and integration of the electronic health record, pathologist workstation, WSS, AP-LIS, and PACS would provide a streamlined workflow for quality patient care.
Primary diagnosis refers to making the final reported pathology diagnosis when reviewing WSI, without first looking at the glass slide. Using WSI for primary diagnosis has been approved in Europe and Canada for several years, and in 2017, the US Food and Drug Administration approved whole slide imaging for primary diagnosis.15 To exemplify the validity of digital pathology performance compared with brightfield microscopy, many validation studies have been performed in all pathology subspecialties with high concordance rates between both methodologies.16–29 These studies have effectively proven that digital pathology is noninferior to conventional glass slide microscopy in terms of diagnostic concordance for primary diagnosis. Furthermore, the College of American Pathologists have put forth guidelines for clinical validation of whole slide imaging.30 Factors included in the validation guidelines will be discussed in another section (see the Clinical validation section). Concordance rates of interobserver and intraobserver variations have been reported with rates similar to brightfield microscopy. LIS integration with WSI has likely increased adoption of DPSs, where pathologists can launch WSI from the LIS and view in WSI viewer software, and subsequently continue reporting of patient pathology in the LIS (ie, LIS-centric DPS). Owing to the proposed digital workflow, using WSI for primary diagnosis could help escalate integration of DPSs with EMRs and other enterprise health care information systems.
Some laboratories have transitioned to a fully digital workflow. Depending on the laboratory infrastructure, the workflows can be described as LIS driven or PACS driven.31 In pathology departments where glass slides continue to dominate the workflow, such physical glass slide workflows are generally AP-LIS driven requiring the integration of stand-alone WSI viewer software. Regional networks with a pressing need for remote digital consultation have instituted hybrid workflows, where digital consultation from remote sites is PACS driven and at larger centers where physical glass signout predominates, there is perpetuation of an AP-LIS-driven workflow. PACS-driven workflows have largely been implemented in Europe, where laboratories all share the same LIS, which enables seamless data and image sharing.
Consultative diagnoses may be referred to any diagnosis being made on patient material after primary diagnosis. Many hospital systems across the nation have requirements for pathology review for incoming treatment of patients. These are commonplace for patients receiving treatment at another hospital system, where hospital mandates require medical diagnoses be confirmed before initiating treatment. This review currently includes packaging, shipment, and reaccessioning of pathology material to the consulting institution. Other reasons for secondary diagnosis are related to diagnostic concerns, either for routine diagnosis, or for intraoperative consultation. Using digital pathology, the WSIs can be digitally shared from one hospital system to another for review before starting treatment. Intrainstitutional consultation is commercially available through dedicated web-based WSI consultation platforms.32 This avoids the inefficient process that is currently in place where pathologists from the primary institution select slides from a case that are being requested for consultation, the slides and possibly tissue blocks are packaged and mailed to the requesting corresponding institution, thereby delivered to the pathology department to be accessioned and assigned to a pathologist for review. Digital pathology for intraoperative consultation also has potential to change current practice paradigms. WSI has been shown effective with intraoperative consultation.33 Pathologists can remotely connect to a secure network where the intraoperative tissue has been prepared on a glass slide and is placed in a dynamic-robotic live streaming WSS.34 Other institutions have provided links or QR codes to hospitals on patient reports for easy access to web-hosted WSI for outside referral. Consultations also commonly occurs within the same institution pathology department. With intradepartmental pathology consultations, DPSs could seamlessly allow for multiple pathologist viewing of cases, or messaging/notification systems to request for a prompt consult. The international pathology landscape is also developing, pathology portals to ingress international consults has been established.35 Furthermore, hospitals and private companies have taken advantage of WSI to offer diagnostic services to underserved countries as well.36 Separate validation studies for diagnosis of WSI as a consultation have been performed and also show decidedly concordant results.37,38
Telepathology allows pathologists to adequately perform their duties without physically existing in the same location they would have otherwise been required to be. Telepathology has the ability to engage pathologists from the gross workstations to the final diagnosis of the slides. With proper technology and implementation, pathologists can interact and review all processes remotely. Gross workstations with integrated cameras can be used for remote streamed viewing of gross specimens for patient care or for instructional training purposes. Another prime example of using digital pathology for telepathology is illustrated for intraoperative consultations. Most hospitals have a recommended time to diagnosis for intraoperative consultations (ie, 20 min). Digital pathology can create efficiencies for off-site pathologists to provide ready access to diagnose the tissue from their off-site workstation. Furthermore, with dynamic-live WSS, the pathologist can remotely navigate and focus the slide in real time while ascertaining a diagnosis. Also used in a similar principle, cytologic rapid on-site evaluations (ROSE) are procedures that usually require a pathologist and/or cytotechnologist to evaluate quantities of cellular tissue. Using these technologies, ROSE procedures no longer require the pathologist to be onsite if cytotechnologists selecting appropriate regions of interest (ROI) to display remotely to the pathologist.39 For remote sites that cannot feasibility host cytotechnologists onsite or on-standby, robotic WSS technologies allow for cytotechnologists to remotely control from a central workstation to render assessments for ROSE procedures.7
Multidisciplinary Diagnostic Management Meetings
Multidisciplinary meetings where physicians discuss patient care and optimal management can also be impacted by digital pathology. Throughout the patient presentations, all relevant medical imaging is shown including radiology, pathology, and any other subspecialty information that may be of clinical relevance. Radiologists can access their enterprise PACS to retrieve all image data from the patient and can easily interchange between connected hospital systems. Digital pathology allows the same archival, storage, and retrieval of pathology-related imaging to be reviewed at the patient conference. Having all the medical imaging digitized, enhances conferences such as tumor boards where patients may opt to attend remotely, or potentially discuss challenging medical cases with other sites in a distributed health network or interinstitutional hospitals. Distributed medical centers have implemented commercially available web-based platforms to allow for pathologist participation in multidisciplinary patient management meetings, where constraints of such meetings being scheduled at distant sites, and where having pathologists expend commuting time to participate is unreasonable.32 These enhancements in pathology workflow have opportunities to provide better patient care and allow for process efficiencies.
The College of American Pathologists published validation guidelines for diagnostic purposes that outlines 12 guidelines.30 All institutions or practices considering the implementation of a digital pathology system for clinical diagnostic purposes must carry out their own validation. Validation for each diagnostic application and use case is necessary and WSI for applications that have not been validated for clinical purposes should not be used. The validation study should closely emulate the real-world environment in which it will be practiced. The entire DPS should be validated collectively instead of validating each subcomponent individually. However, revalidation of the entire system is required whenever a significant change is made to any component of the DPS. Documentation should be maintained recording the methods, measurements, and final approval of validation for the system to be used in clinical practice. A pathologist adequately trained to use the WSI system must be involved in the validation process. The guidelines further describe that validation of DPSs need only involve specific types of specimens and preparations, but not specific tissues, diseases, microscopic changes or diagnoses. Approximately 100 cases that reflect the spectrum and complexity of specimen types and diagnoses should be included in the validation. Randomizing the glass slide and WSI review should be performed. A washout period of at least 2 weeks should occur between viewing digital and glass slides. The intended tissue present on the slide should be included in the validation process (ie, all tissue, or specified area) to be scanned and included in the digital image. Intraobserver diagnostic concordance between digital and glass slides for the same observer should be measured.30 Interobserver variability has been shown to reflect diagnostic interpretation differences as opposed to differences in using different technologies, as it not recommended in the guidelines.40
Using WSI for educational purposes provides access to high quality, reliable data. Pathology trainees learn by (re)viewing slides, creating algorithms of differential diagnoses, and rendering an accurate diagnosis. Aside from routine clinical cases, study sets of glass slides can be organized and stored for the slide’s life duration. Some challenges to using glass slides for education necessitates the trainee to have a microscope at the given time they want to review the study set, and only allow for the 1 trainee to view 1 slide at a time. This can be circumvented by using a multiheaded microscope where multiple trainees can view 1 slide simultaneously, which is currently the best method for group teaching. Also, the reagents (eg, H&E) are chemicals that may fade over time. With digital pathology, WSI are not affected by slide breakage, reagent fading, or shipment tracking/loss; they can be organized into atlases for ready access review. Multiple slides can also be simultaneously viewed, making contrasts of comparable diagnoses easier to distinguish. In addition, multiple trainees can view the same WSI for group educational sessions, providing reproducibility for each FOV of a given slide. A prime example of where this matters greatly is during standardized examinations. Each year, the American Board of Pathology has increasingly used WSI during pathology board examinations, this ensures reproducible images and ensures each candidate is rendering a diagnosis on the same WSI. The same principles can be applied for Continuing Medical Education. As physicians are required to complete Continuing Medical Education each year postgraduation, courses may offer WSI as part of their web-based modules. WSI grant potential for ready access to slides for (inter)national presentations, conferences, and journal publications alike.
Daily routine workflow includes efficient management of resources, turnaround time, materials, and personnel; pathology administration is a major component of the laboratory. Quality assurance of reagent staining, slide preparations, and digital imaging; remote access allows for conducting real-time quality evaluations. Quality assessments can be performed online or in the anatomic pathology LIS to decrease turnaround time. Having ready access to WSI can perhaps help standardize a specific reagent concentration in the department as a step toward color calibration for the field as a whole. Digital imaging alleviates visual trending biases and also provides the possibility to set a normalized color calibration, or preferred color settings, across all WSI. Also, retrieval of prior patient pathology digitally archived can allay frustrations with pulling prior glass slides from a physical slide archive (ie, offsite storage), and decrease TAT in rendering patient diagnoses.
Exemplified in the regulatory requirements that oblige cytology pap smears to a 5-year look-back such that any diagnosis equal to or worse than a high-grade squamous intraepithelial lesion requires all negative pap smears in the last 5 years to be reviewed. Altogether, digital pathology motions to provide novel and innovative quality metrics for the current analog paradigm. Intradepartmental case rereview is integral to an anatomic pathology quality assurance system. There are several advantages of WSI over rereview of glass slides. These advantages stem from the ready access and easy sharing of WSI compared with glass slides. AP-LISs currently have the capability to prospectively flag a specified percentage of cases for rereview before finalization. WSI can be used for prospective QA review and help to identify and correct errors in a time efficient manner before they reach the patient. This is of particular benefit to institutions that incorporate digital workflows, where QA functionalities are built into their image management systems. Many practices also require a second pathologist to review cases with newly diagnosed malignancies or medico-legal review as part of their QA plan. The use of digital pathology systems allows for pathologists to identify these critical cases for review and share them instantly to the worklist of a colleague.
Hitherto, pathologists have relied on their trained eye and clinical expertise for diagnosis and disease classification. Having WSI allows the potential to create more objective measurements. Digital pathology research, in regards to image analysis, has shown evidence of clinical importance. Novel tumor classifications based on epithelium to stroma ratios may be more empirically calculated by software image analysis tools.41 Or variations in the tumor microenvironment to stratify prognosis based on subtypes of lymphocytes.42 Research journals with online publications may allow for web-hosted WSI to be included in part of publications with WSI figures embedded in the journal publication including annotations denoting specified regions of interest. Cancer registries and other big data resources solicit WSI to build cohorts with patient-associated metadata. These workflows and systems can be used for clinical trials for streamlined pathology review. These registries could amass voluminous clinical data for prognostic and medical treatment discoveries using machine learning models.
Image analysis has unprecedented potential for increasing the value of digital pathology. It references applications or algorithms to enable new interpretation of WSI, including associated clinical data, as well as computer-assisted diagnosis. Feature selection and extraction is one such method to image analysis can be leveraged for classification. Spatial relationships of various defined cellular and stromal components, such as nuclei, cytoplasm, or cell membranes can provide new information for computational algorithms to find clinical significance. A network of features can be aggregated to ultimately provide tissue classification. Image-based searching (ie, content-based image retrieval), allows for searching of similar images based on pixel information. The WSI analysis is not limited solely to H&E staining, but to the armamentarium of immunohistochemical or special stains available that already exist to pathologists to assist for diagnostic purposes. Within the past several years, image analysis has approved algorithms to automatically quantify and score immunohistochemical studies.43 There are US Food and Drug Administration approved biomarkers available for quantification using the specified reagents and software. This permits a more standardized and objective cellular quantification. This technology is used to aid the pathologist quantification and classification of cells based on pattern, shape, size, and pixel color information. This is particularly applicable where multiplex immunofluorescent imaging with biomarkers will require computational assessment for quantitation, and where the traditional manual observation of subjective human assessment proves inadequate. Other applications include automation or efficiency in detection of microorganisms (eg, acid fast bacilli), where pathologists spend significant time screening each glass slide.44
Machine learning applications in medicine exist in radiology, ophthalmology, and cardiology, among other image-based specialties, however digital images in pathology are more complex and of larger size. In recent years, advancements in computing power, data, and processes have tremendously evolved. The application of machine learning and deep learning technologies in digital pathology have recently emerged and include clinical grade developments.45 These tools can be used for clinical decision support where pretrained models are deployed to assist pathologists in detection, segmentation, or classification of patient pathology. In addition, other applications include using machine learning for education, and quality assurance.46 One of the key driving forces for machine learning applications is to use the pixel data embedded within whole slide images and integrate with other clinico-pathologic or genomic data to potentially establish novel digital biomarkers for diagnosis or prognostication. Pathology one of the largest image-based medical specialties is transitioning to a digital workflow, whereby subsequently can take advantage of related machine learning applications. These changes will offer pathologists a new standard of care for patients, however there are still regulatory, technical, ethical, and legal considerations. Enabling digital pathology with machine learning will also offer new research and educational insights that can pave the way for personalized medicine in an unprecedented manner to enhance patient care.
Nondestructive tissue analysis is available with imaging techniques such as optical coherence tomography or microscopy with ultraviolet sectioning excitation can allow rapid, superficial imaging of tissue.47 Image data can even be reconstructed into volumetric data sets to create histopathologic 3D tissue blocks. These methods bypass traditional histology, where no sectioning or staining of tissue is required. Further studies are in progress in pathology as well as other clinical fields to further develop these technologies.
Using specialized cameras, a multispectral image is created by capturing the same image several times. Each image is acquired using a different filter with multiple sensors, each sensitive to a specific wavelength of color. Multispectral imaging requires the use of specialized cameras, hardware, and software, different from the RGB-only WSI scanners. The hardware consists of optical filters that control liquid crystals to transmit a specific wavelength of light and exclude all other wavelengths. These liquid crystal tunable filters and monochrome CCD sensors capture images that can be used in the brightfield-based or fluorescence-based microscopy. The software can further analyze images with different spectral signatures by decoding each of the multiple assays for different classifications. Each class can potentially be used to discriminate between cell or tissue types. This process of unmixing each band into their individual spectra is made possible by capturing each antibody stain spectra individually, such that each stain can be directly measured and analyzed interchangeably with other stains independently. Using computational algorithms such as linear or nonlinear unmixing, the differences of each spectra can be distinguished and used for analysis. Potential in classifying dermatological diseases based on differences of the frequency of unique classified spectra, or by differing ratios of classified spectra has been shown.48 Spectral libraries can be configured to maximize optical magnification and increase spectral diversity and ultimately provide accurate classification. Development of these innovative technologies will support morphologic assessment and develop a multidisciplinary imaging diagnostic modality.
Digital pathology and cloud-based systems have enabled users unprecedented interconnectivity and progress. As DPS continue to mature, innovative technology with increased functionality have allowed researchers to develop color calibration slide to promote color standardization in WSI.49,50 Understanding quality control metrics and their impact on the quality of WSI is important to consider their interferences with image analysis algorithms.51–53 Software to stitch and align WSI are available,54 including generating 3D WSI. New advances can construct 3D histopathology images from serial scans of tissue blocks. Volumetric pathology, or 3D microscopy, can reveal novel areas of interest in histology and provide new positional information of reconstructed volumetric data.55,56 Integration with existing information systems, such as molecular laboratory information management system, solutions are being created that support the integration of heterogeneous data sets including clinical, next-generation sequencing, and imaging data.57 As these technologies continue to be developed, they will shape pathology workflow and patient care in the digital era.
The sophistication of the digital pathology ecosystem offers novel clinical, nonclinical, and research applications. The DPSs include varied hardware and software components based on their intended use. Emerging tools using WSI are growing and are subsequently increasing their associated applications. WSI have the potential to improve diagnostic accuracy, workflow efficiency, balance workloads, integrate siloed HIS, and also provide a profitable return on investment.58 WSI standardization may provide higher quality assessment and enable image analysis tools to be developed and validated for use. These systems and features will facilitate pathologists to embrace these technologies and will transform patient care in a modern era of pathology.
1. Parwani AV, Feldman M, Balis U, et alPantanowitz L, Tuthill MJ, Balis U. Digital imaging. Pathology Informatics: Theory and Practice. Canada: ASCP; 2012:231–260.
2. Ferreira R, Moon B, Humphries J, et al. The virtual microscope. Proc AMIA Ann Fall Symp. 1997:449–453.
3. Pantanowitz L, Valenstein PN, Evans AJ, et al. Review of the current state of whole slide imaging
in pathology. J Pathol Inform. 2011;2:36.
4. Armstrong-Smith I. Facing the Digital Future of Pathology. 2014. Available at: https://thepathologist.com/issues/0114/facing-the-digital-future-of-pathology
. Accessed August 29, 2018.
5. Weinstein RS, Descour MR, Liang C, et al. An array microscope for ultrarapid virtual slide processing and telepathology. Design, fabrication, and validation study. Hum Pathol. 2004;35:1303–1314.
6. Agarwal S, Zhao L, Zhang R, et al. Facetime validation study: low-cost streaming video for cytology adequacy assessment. Cancer Cytopathol. 2016;124:213–220.
7. Sirintrapun SJ, Rudomina D, Mazzella A, et al. Robotic telecytology for remote cytologic evaluation without an on-site cytotechnologist or cytopathologist: an active quality assessment and experience of over 400 cases. J Pathol Inform. 2017;8:35.
8. Rojo MG, Garcia GB, Mateos CP, et al. Critical comparison of 31 commercially available digital slide systems in pathology. Int J Surg Pathol. 2006;14:285–305.
9. Goode A, Gilbert B, Harkes J, et al. OpenSlide: a vendor-neutral software foundation for digital pathology
. J Pathol Inform. 2013;4:27.
10. Goldberg IG, Allan C, Burel JM, et al. The Open Microscopy Environment (OME) Data Model and XML file: open tools for informatics and quantitative analysis in biological imaging. Genome Biol. 2005;6:R47.
11. Clunie D, Hosseinzadeh D, Wintell M, et al. Digital imaging and communications in medicine whole slide imaging
connectathon at digital pathology
association pathology visions 2017. J Pathol Inform. 2018;9:6.
12. Helin H, Tolonen T, Ylinen O, et al. Optimized JPEG 2000 compression for efficient storage of histopathological whole-slide images. J Pathol Inform. 2018;9:20.
13. Molin J, Lundström C, Fjeld M. A comparative study of input devices for digital slide navigation. J Pathol Inform. 2015;6:7.
14. Krupinski EA, Comas M, Gallego LG. On behalf of the GISMAR Group. A new software platform to improve multidisciplinary tumor board workflows and user satisfaction: a pilot study. J Pathol Inform. 2018;9:26.
15. US Food and Drug Administration. FDA News Release: FDA allows marketing of first whole slide imaging
system for digital pathology
. 2017. Available at: https://www.fda.gov/drugs/informationondrugs/approveddrugs/ucm553358.htm
. Accessed October 18, 2018.
16. Al-Janabi S, Huisman A, Nap M, et al. Whole slide images as a platform for initial diagnostics in histopathology in a medium-sized routine laboratory. J Clin Pathol. 2012;65:1107–1111.
17. Bauer TW, Schoenfield L, Slaw RJ, et al. Validation of whole slide imaging
for primary diagnosis in surgical pathology. Arch Pathol Lab Med. 2013;137:518–524.
18. Brunelli M, Beccari S, Colombari R, et al. iPathology cockpit diagnostic station: validation according to College of American Pathologists Pathology and Laboratory Quality Center recommendation at the Hospital Trust and University of Verona. Diagn Pathol. 2014;9(suppl 1):S12.
19. Buck TP, Dilorio R, Havrilla L, et al. Validation of a whole slide imaging
system for primary diagnosis in surgical pathology: a community hospital experience. J Pathol Inform. 2014;1:43.
20. Campbell WS, Lele SM, West WW, et al. Concordance between whole-slide imaging and light microscopy for routine surgical pathology. Hum Pathol. 2012;43:1739–1744.
21. Cheng CL, Azhar R, Sng SH, et al. Enabling digital pathology
in the diagnostic setting: navigating through the implementation journey in an academic medical centre. J Clin Pathol. 2016;69:784–792.
22. Fonyad L, Krenac T, Nagy P, et al. Validation of diagnostic accuracy using digital slides in routine histopathology. Diagn Pathol. 2012;7:35.
23. Gilbertson JR, Ho J, Anthony L, et al. Primary histologic diagnosis using automated whole slide imaging
: a validation study. BMC Clin Pathol. 2006;6:4–19.
24. Goacher E, Randell R, Williams B, et al. The diagnostic concordance of whole slide imaging
and light microscopy. Arch Pathol Lab Med. 2017;141:151–161.
25. Houghton JP, Ervine AJ, Kenny SL, et al. Concordance between digital pathology
and light microscopy in general surgical pathology: a pilot study of 100 cases. J Clin Pathol. 2014;67:1052–1055.
26. Jukic DM, Drogowski LM, Martina J, et al. Clinical examination and validation of primary diagnosis in anatomic pathology using whole slide digital images. Arch Pathol Lab Med. 2011;135:372–378.
27. Mukhopadhyay S, Feldman MD, Abels E, et al. Whole slide imaging
versus microscopy for primary diagnosis in surgical pathology: a multicenter blinded randomized noninferiority study of 1992 cases (Pivotal study). Am J Surg Pathol. 2018;42:39–52.
28. Pagni F, Bono F, Di Bella C, et al. Virtual surgical pathology in underdeveloped countries: the Zambia project. Arch Pathol Lab Med. 2011;135:215–219.
29. Snead DR, Tsang YW, Meskiri A, et al. Validation of digital pathology
imaging for primary histopathological diagnosis. Histopathology. 2016;68:1063–1072.
30. Pantanowitz L, Sinard JH, Henricks WH, et al. Validating whole slide imaging
for diagnostic purposes in pathology. Guideline from the College of American Pathologists Pathology and Laboratory Quality Center. Arch Pathol Lab Med. 2013;137:1710–1722.
31. Baidoshvili A, Bucur A, van Leeuwen J, et al. Evaluating the benefits of digital pathology
implementation: time savings in laboratory logistics. Histopathology. 2018;73:784–794.
32. Jones NC, Nazarian RM, Duncan LM, et al. Interinstitutional whole slide imaging
teleconsultation service development: assessment using internal training and clinical consultation cases. Arch Pathol Lab Med. 2015;139:627–635.
33. Huang Y, Lei Y, Wang Q, et al. Telepathology consultation for frozen section diagnosis in China. Diagn Pathol. 2018;13:29.
34. Evans AJ, Chetty R, Clarke BA, et al. Primary frozen section diagnosis by robotic microscopy and virtual slide telepathology: the University Health Network experience. Semin Diagn Pathol. 2009;26:165–176.
35. Zhao C, Wu T, Ding X, et al. International telepathology consultation: three years of experience between the University of Pittsburgh Medical Center and KingMed Diagnostics in China. J Pathol Inform. 2015;6:63.
36. Boggan JC, Walmer DK, Henderson G, et al. Vaginal self-sampling for HPV infection as a primary cervical cancer screening tool in a Haitian population. Sex Transm Dis. 2015;42:655–659.
37. Molnar B, Berczi L, Diczhazy C, et al. Digital slide and virtual microscopy based routine and telepathology evaluation of routine gastrointestinal biopsy specimens. J Clin Pathol. 2003;56:433–438.
38. Perron E, Louahlia S, Nadeau L, et al. Telepathology for intraoperative consultations and expert opinions: the experience of the eastern Quebec telepathology network. Arch Pathol Lab Med. 2014;138:1223–1228.
39. Lin O, Rudomina D, Feratovic R, et al. Rapid on-site evaluation using telecytology: a major cancer center experience. Diagn Cytopathol. 2019;47:15–19.
40. Fallon MA, Wilbur DC, Prasad M. Ovarian frozen section diagnosis: use of whole-slide imaging shows excellent correlation between virtual slide and original interpretations in a large series of cases. Arch Pathol Lab Med. 2010;134:1020–1023.
41. Linder N, Konsti J, Turkki R, et al. Identification of tumor epithelium and stroma in tissue microarrays using texture analysis. Diagn Pathol. 2012;7:22.
42. Galon J, Pagès F, Marincola FM, et al. Cancer classification using the Immunoscore: a worldwide task force. J Transl Med. 2012;10:205.
43. Stålhammar G, Fuentes Martinez N, Lippert M, et al. Digital image analysis outperforms manual biomarker assessment in breast cancer. Mod Pathol. 2016;29:318–329.
44. Tadrous PJ. Computer-assisted screening of Ziehl-Neelsen-stained tissue for mycobacteria. Algorithm design and preliminary studies on 2,000 images. Am J Clin Pathol. 2010;133:849–858.
45. Campanella G, Hanna MG, Geneslaw L, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nat Med. 2019;25:1301–1309.
46. Niazi MKK, Parwani AV, Gurcan MN. Digital pathology
and artificial intelligence. Lancet Oncol. 2019;20:e253–e261.
47. Qorbani A, Fereidouni F, Levenson R, et al. Microscopy with ultraviolet surface excitation (MUSE): a novel approach to real-time inexpensive slide-free dermatopathology. J Cutan Pathol. 2018;45:498–503.
48. Kim S, Cho D, Kim J, et al. Smartphone-based multispectral imaging: system development and potential for mobile skin diagnosis. Biomed Opt Express. 2016;7:5294–5307.
49. Martina JD, Simmons C, Jukic DM. High-definition hematoxylin and eosin staining in a transition to digital pathology
. J Pathol Inform. 2011;2:45.
50. Bautista PA, Hashimoto N, Yagi Y. Color standardization in whole slide imaging
using a color calibration slide. J Pathol Inform. 2014;5:4.
51. Campanella G, Rajanna AR, Corsale L, et al. Towards machine learned quality control: a benchmark for sharpness quantification in digital pathology
. Comput Med Imaging Graph. 2018;65:142–151.
52. Yagi Y, Gilbertson J. A relationship between slide quality and image quality in whole slide imaging
. Diagn Pathol. 2008;3:S12.
53. Bautista PA, Yagi Y. Improving the visualization and detection of tissue folds in whole slide images through color enhancement. J Pathol Inform. 2010;1:25.
54. Toth RJ, Shih N, Tomaszewski JE, et al. Histostitcher™: an informatics software platform for reconstructing whole-mount prostate histology using the extensible imaging platform framework. J Pathol Inform. 2014;5:8.
55. Hanna MG, Ahmed I, Nine J, et al. Augmented reality technology
using Microsoft HoloLens in anatomic pathology. Arch Pathol Lab Med. 2018;142:638–644.
56. Roberts N, Magee D, Song Y, et al. Toward routine use of 3D histopathology as a research tool. Am J Pathol. 2012;180:1835–1842.
57. Dander A, Baldauf M, Sperk M, et al. Personalized oncology suite: integrating next-generation sequencing data and whole-slide bioimages. BMC Bioinformatics. 2014;15:306.
58. Ho J, Ahlers SM, Stratman C, et al. Can digital pathology
result in cost savings? A financial projection for digital pathology
implementation at a large integrated health care organization. J Pathol Inform. 2014;5:33.