Digital Pathology

From Clinfowiki
Jump to: navigation, search

Digital Pathology is a broad term defined as a dynamic image-based environment that enables the acquisition, management and interpretation of pathology information generated from a glass slide. [1]. Or simply it is digitization of pathology. Whole slide imaging is commonly equated to digital pathology but digital pathology more broadly includes the capture of pathology slides by a simple camera (static and video) on a microscope, or robotic microscopy (the use of a robot to move image capturing system real time). However, some may argue that photographic images of gross specimens with integration into laboratory information systems can be considered digital pathology [2].


Digital Pathology is a much talked about field in pathology informatics. In the late 90’s and early 2000’s as digital cameras became increasingly used, microscopy photography of pathology slides became increasingly popular. It was also around this time that whole slide scanners (WSI) were first introduced [3].

Major advantages and justifications for Digital Pathology include:

  • Education – Digital slides allow for greater access and sharing of complex pathology cases to pathologists around the world. Instructors can highlight regions of interest and many slide repositories are available on the web [4]. Prior to digital pathology complex cases would require recuts and multiple recuts may result in loss of the cells of interest and chronic exposure to light may dim the intensity and contrast of stains on glass slides.
  • Research – Digital slides potentially allow automated detection of cells of interest to improve accuracy and reproducibility in the detection of abnormal cells.
  • Telepathology – Expert pathologists will be able to make rapid frozen section diagnoses outside the hospital. Consultation of difficult cases may be more streamlined and have a quicker turnaround time.
  • Medical Record Integration – Currently much of the patient’s anatomical pathology data resides outside of the medical record. Increase integration of pathology images will allow greater transparency of pathology results to other clinicians and even patient’s themselves much like radiology images today.

Currently the largest barriers to implementation include: [5]

  • Cost – WSI scanners can run upwards of at least 200 thousand dollars and there is significant amount of cost associated with the storage and retrieval of whole slide images.
  • Regulatory issues – It has only been 1 year since a WSI system has been approved for use by the FDA for primary diagnosis.
  • Perceptions of inferiority – Many pathologists are more familiar with the workflow of signing out glass slides and adjustments to the digital pathology workflow must be made. However, studies have demonstrated it is realistically possible to achieve efficient use of digital pathology for sign-out [6] [7]. However, for cytology specimens and hematopathology specimens that require viewing multi-dimensional cells in 60X magnification there is still superiority in light microscopy at this time.
  • Lack of imaging standards [8].

Digital Cameras

Unlike WSI, digital cameras are commonly used in the pathology lab. However, they are not used for clinical sign-out. For the most part digital cameras have been used for photography of gross specimens and the capture of single images of a pathology slide from a camera attached to the microscope. Microscope cameras are relatively cheap to implement and are predominately used for teaching and/or preparing of presentations or slides for multidisciplinary tumor boards. Video cameras attached to the microscope are commonly used in tumor board and teaching settings as well. Gross photography on the other hand can be implement into the LIS for record keeping purposes as well. Proper protective measures from the outside environment need to be incorporated with most implementations using a shell to shield the camera. Hand-free controls and/or remote controls that easily can be easily cleaned are utilized to further avoid contamination of outside environment like blood or body fluids and avoid significant disruption to workflow (since without it the user would have to go from a dirty environment to a clean environment every time a picture is needed).

Whole Slide Imaging (WSI) Barriers

While WSI holds much promise there has been significant barriers to implementation

WSI Scanning and Image Formation

Early slide scanners could not load many slides at a time and scanned at very slow speeds. Due to this reason, many argued in the beginning that WSI is too disruptive and costly to implement [9]. A large Pathology laboratory may generate up to 2400 slides per day. In order to process those number of slides with one scanner, the scanning speed must take less than a minute a slide and early scanners were significantly slower. The new FDA approved Philips Intellisite Pathology Solution scans a 15 x 15 mm area at 40X in 60 seconds and hold 300 slides at a time [10]. WSI scanners move the glass slide in XY axes and sometimes in the Z axes in order to capture the entire slide. Individual images are stitched together in a single large slide image. In order to reduce file size most modern whole slide imagers scan for regions of interest (ie where the tissue is located in the slide). In most surgical pathology slides the stained tissue only covers a small portion of the slide. In contrast, in cytology smears may cover the entire glass slide and may require capture of multiple levels (Z axes) and therefore have a significant increase in both size and time to digitization. WSI of common specimens range up to 1-4 GB per slide and with busy laboratories outputting at least 2400 slides a day this equates to potentially 9.6 TB a day. Therefore, the storage costs of maintaining a virtual repository of pathology slides is quite significant.


Interoperability of digital pathology images is difficult because no true universal format exists for both user support or for archiving. Many modern slide scanner manufactures have their own proprietary file formats with significant variance in compression techniques that affects the quality output from these scanners. Many argue for this reason whole slide images should be saved in the DICOM format. DICOM has a standard for Pathology images which was developed by working group 26 that was only approved in 2010. Slide scanner vendors have slowly been incorporating this format with their provided image software. Last October “Connectathon” shows were put on where vendors (Philips, Ventana, Leica) proved they had interoperability by scanning and importing the resulting images using the DICOM standard into an agreed upon PACS system ahead of conference [11].

Regulatory Issues

There also have been significant regulatory issues concerning digital pathology which has slowed incorporation into clinical workflows. Light microscopy, is registered as a class I medical device and therefore does not need FDA approval to be sold in the healthcare setting [12]. However, in 2009 the FDA held a meeting to discuss the replacement of light microscopy by WSI for primary diagnosis. They debated making WSI a class II device requiring a 510K approval or a class III device requiring premarket approval clinical trials, with the latter being the final conclusion after a few years of discussion [13]. The primary diagnosis verbiage is emphasized to distinguish it from WSI of immunohistochemical stains. Images of these types are used to quantify the expression of prognostically important molecules but do not have diagnostic implications. Given the perceived reduced risk, these scanners and the image analysis algorithms associated with them classified as Class II devices. It took the FDA another 5 years to provide draft recommendations in 2014 on how to perform an assessment of WSI devices, which wasn’t finalized till 2016 [14]. Because of this uncertainty, regulatory approval of the first whole-slide scanner for diagnostic use didn't occur until 2017, when the FDA approved the Philips IntelliSite Pathology Solution (PIPS). The FDA approved PIPS as a class II medical device, setting the stage for an easier road to approval for other whole-slide scanners. With this precedent, the FDA has reduced the regulatory barrier to WSI implementation.

Interface with LIS

Interfacing the LIS with pathology images is necessary for clinical workflows. However, full integration with the LIS is not entirely necessary for pathology images and may be difficult with the current state of laboratory information systems. A modular or integrative approach may be used. [15] In the modular approach, images are stored outside of the LIS and a middleware solution is used to connect the image repository to the LIS. This has an increased flexibility in the sharing and viewing of the images. Vendor neutral archives (VNA) are getting increased traction in the medical field today and the modular approach may allow the incorporation into these archives easier.

WSI Image Applications

Basic Image Analysis

Digital Pathology has great potential in automating the detection of cells of interest. Quantification of positively staining cells for biomarkers in immunohistochemical (IHC) slides can be monotonous with significant inter-observer variability in pathology. Images analysis of these slides holds much promise and may allow for increased accuracy, more reproducible results, increased automation and reduce time consumption by pathologists. Image analysis of HER2 stain slides breast cancer has the longest history of utilization, and therefore has clear formalized guidelines governing its application for us. Studies have demonstrating superiority in manual assessment by image analysis [16]. Other biomarkers with FDA cleared image analysis algorithms include ER, PR, Ki67, cytokeratin, and p53.

Image analysis algorithms mentioned above require 510(k) FDA clearance; however, there has been a recent drive to create laboratory developed tests (LDTs) utilizing image analysis. LDTs are not FDA-approved or evaluated, and they're generally not used widely. Laboratory regulations and accreditation standards represent a huge barrier to the use of LDTs in laboratories in the United States. Advanced techniques involving the detection of malignant cells/patterns in H&E slides are not currently used clinically, but this is the focus of many contemporary research endeavors. Simple image analysis techniques are usually insufficient for H&E and artificial intelligence and deep learning techniques are commonly incorporated.

Convolutional Neural Networks

Convolutional Neural Networks can be manipulated to teach a computer how to assess an image. The process generally involves training a computer by showing it numerous images and telling it what diagnosis is represented by each image. Complicated mathematical models convert images into numbers and matrices, and the computer compares and analyzes these values to develop a framework with which to classify new images it has never seen before [17]. The examples of use cases for these types of image analysis networks are growing exponentially in the literature. Algorithms for detection of cancer in cervical biopsies [18]and the grading of prostate cancer has recently been reported [19]. Algorithms for determining outcomes based on image analysis have been explored for brain cancer [20] and breast cancer [21], at a minimum.


Telepathology is a specific application of WSI that allows people to view pathology slides from around the globe. Historically, pathology slides have been shared/distributed by mailing glass slides from one location to another. With the advent of microscope cameras and the internet, telepathology allowed for digital pathology images to be transmitted. However, whole-slide images allow for a more optimal form of telepathology, because they allow the recipient to view an entire slide at multiple levels of magnification. With WSI, telepathology can occur from across the room, across the hospital, or across the globe. Aside from diagnosis, telepathology can also assist with quality assurance, research, and education.

Frozen section evaluation is one example of how telepathology allows for the provision of care to underserved and rural areas. Multiple pathologists in different locations, including specialists/experts, can view frozen section slides and allow for the most appropriate surgical care for the patient. One study comparing telepathology to the traditional method of viewing frozen section slides (direct, in-person microscopy) found that there was no statistical difference in diagnostic accuracy between the two methods [22]. Telepathology is also helpful in determining the adequacy of cytology specimens, such that remote pathologists can determine if the sampling is adequate while the patient is still in the office or procedure suite.

Select Digital Pathology References and Resources

[College of American Pathologists (CAP) Website; Digital Pathology Resource Guide] [/ Digital Pathology Association] [Digital Pathology: A 1st anniversary report card ] [Iowa Digital Slide Box]

Submitted by Thomas Schneider and Valerie Lockhart (Telepathology, Convolutional Neural Networks, edits and additions to other sections)