Digital Pathology 101 – Pros and cons of making the switch to digital workflow

Motic Digital Pathology Launches MoticFlow
Image Credit: Motic

Pathologists have traditionally worked with glass slides and conventional microscopes, but the digital revolution has allowed a transformation from glass to digital slide. This process entails capturing the entire glass slide with a scanning device to obtain a high-resolution whole slide image (WSI) for diagnostic, educational, or research purposes. This slide can then be viewed, managed, shared, analysed, and interpreted. The integration of all the pathological information generated from digitized glass slides is encompassed within the technical and scientific field now known as ‘digital pathology’.

There are some clear differences between the two modalities of slide, such as the traditional round field of view from a microscope versus the rectangular field of view provided by image viewing software. Microscopes obviously involve manual manipulation whereas the digital method uses computer controls for slide navigation. Pathologists should however be aware that although a WSI is designed to be a precise replica of a glass slide, it may not be an exact copy in every possible way. During the scanning process computer-controlled parameters such as resolution, contrast, brightness, color rendition, or depth of field can contribute to the information which is captured from the slide and then presented for assessment to the pathologist.

So, can we consider that digital and glass slides are really equivalent? This topic has been explored in several validation studies that examined intra-observer concordance on diagnosis when comparing WSI with traditional glass slides. Although some authors have reported differences in concordance between the two modalities of slides in some specific diagnostic subgroups, most of the published literature has concluded that digital slides are at least equivalent to glass slides for rendering diagnosis (1).

However, digital pathology offers many benefits and advantages that cannot easily be achieved with glass slides. It enables electronic transfer of WSI from the laboratory to the pathologist, wherever he or she may be. Not all hospitals have a pathologist on site, and hence digital pathology in this case enables remote and rapid assessment of their tissue samples. This may reduce the time to diagnosis, enabling the rapid referral of the patient to a specialist. Moreover, the WSI enables instant sharing among peers, increases the access to second opinions, and even facilitates the work of multidisciplinary teams. Digital pathology can also reduce the risk of patient misidentification and, since the WSI becomes a part of the patient’s electronic medical record, they can easily be accessed and are less likely to go astray.

Notwithstanding these key benefits of digital pathology, glass microscope slides are still proving to be important and necessary. Before digitization, the tissue samples still need to be fixed, sectioned, stained and mounted. Moreover, implementation of digital pathology in the laboratory requires a significant investment in hardware, software, IT infrastructure and training. It may also necessitate an overhaul of IT protocols, laboratory workflows and an adjustment to storage and backup planning.

Some of the technical considerations include how to identify the correct specimen during the scanning process either manually or with a barcoding system. One should also be aware that WSI systems perform at their best under peak optical conditions and quality standards must be put in place during the lab sectioning and staining processes to optimise the available scanning area and the ability of the instrument to correctly focus on the target tissue section. Moreover, retention and archiving policies must be established for both glass slides and digital images and there may be patient confidentiality and legal issues to consider when sending images digitally to colleagues in other countries.

For pathologists, the transition to digital pathology requires a not unsubstantial training process during which they must learn the new digital pathology workflow, familiarize themselves with the software and to diagnose from a digital on-screen image.  Pathologists must also become adept at manipulating the image files themselves, learning how to access the patient documentation and then to open, display, navigate and zoom the WSI. It takes time and experience to become familiar with the more subtle elements of the slide scanning process such as the identification of scanning artifacts and the lack of focal plane.

To ensure that digital pathology is safe and effective it is recommended that the laboratory performs a full validation and verification process prior to implementation. Validation should be performed by the pathologist via comparison of diagnosis from both glass and digital slides on the same set of cases. The goal of the process should be to prove that when using digital pathology rather than traditional microscopic methods, the diagnostic standards, patient safety and clinical effectiveness of the laboratory are maintained.

Although requiring substantial time and financial investment, the implementation of digital pathology has been shown to save time, improve workflow and to lead to better utilisation of resources with higher efficiency and greater levels of collaboration. However, the use of digital pathology opens up other possibilities for expanding analytical and diagnostic potential which go far beyond the simple concept of digitisation and sharing of whole slide images.

During the last decade there have been huge advances in the development of image analysis software applications that can automatically recognise and measure structures, quantify biomarkers or, for example, detect hot spots of biomarker positive cells in the image. These image analysis tools allow pathologists and researchers to automate, to some extent, the time-consuming process of sample evaluation and analysis, Moreover, the detection capabilities of image analysis software, when dealing with subtle changes in staining intensity or tissue morphology, can far surpass the capabilities of the human eye.

Even more advanced digital pathology applications come in the form of artificial intelligence (using machine learning) and neural networks. Artificial intelligence is the ability of computer systems to interpret and learn from external data and then to apply these ‘learned’ behaviours to perform specific tasks that normally require human intelligence. Artificial intelligence uses algorithms, based on an array of instructions that can be trained with data sets. Algorithms when trained with WSI sample sets can assist with identifying patterns in tissues, grading tumors, detecting regions of interest or could even be leveraged to perform automated pre-screening and diagnostics.

This new paradigm of analysis certainly has the potential to improve both the speed and the accuracy of diagnosis. However, AI based algorithms developed from machine learning are not yet approved for use in diagnostic practice and there are only a few (though an increasing number) of standard image analysis algorithms which have been cleared by the FDA for clinical use.

Artificial intelligence tools have been developed for in vitro prostate cancer detection and have also shown an ability to detect other cancers such as metastases within lymph nodes and malignancies in breast and cervical cancers. Artificial intelligence is also being applied to immunohistochemistry analysis and some image analysis algorithms having already received FDA clearance for the assessment of pathological features such as the expression of ER, PR, HER-2 and Ki-67 in breast cancer..

The number of imaging tools cleared or approved by the FDA for companion diagnostics is also growing rapidly. Some examples of the predictive biomarkers that can be detected with these devices include PDL-1 in several type of cancer, ALK in non-small cell lung cancer, c-KIT receptor in gastrointestinal stromal tumour. Also, genetic alterations such as mutations, insertions, deletions or rearrangements in several genes (EGFR, BRCA1/2, KRAS, PIK3CA, IDH1/2) in colorectal cancer or breast cancer.

The potential applications of artificial intelligence in digital pathology are extensive and include the identification of key morphological and sub-cellular features or the counting of large numbers of cells or cells of a specific type. Some studies have highlighted the fact that artificial intelligence can detect certain cell characteristics, such as mitoses, with the same accuracy as expert pathologists, and may even exceed the Pathologist in speed and efficiency.

Certainly, image analysis and artificial intelligence will play a major role in the future advancement of the field and we can only expect more and more tissue pathology laboratories to make the move to digital technology as the benefits of digitisation continue to accumulate and start to outweigh the substantial investments required.


References:

  1. Ho, J., & Pantanowitz, L. (2017) Making pathology diagnoses with glass or digital slides: Which modality is inferior? J Pathol Inform, 8:14
  2. Cross, S., Furness, P., Igali, L., Snead, D., & Treanor, D. (2018). Best practice recommendations for implementing digital pathology. The Royal College of Pathologists. London. Pp 38.
  3. Neff Newitt, V. (2019) Batter than glass slides? Digital pathology’s challenge. CAP TODAY. September 2019. Available at: https://www.captodayonline.com/better-than-glass-slides-digital-pathologys-challenge/
  4. Moxley-Wyles, B., Colling, R., & Verrill, C. (2020) Artificial intelligence in pathology: an overview. Diagnostic Histopathology, 26:11
  5. Food and Drug Administration. List of cleared or approved companion diagnostic devices (in vitro and imaging tools). Available at: https://www.fda.gov/medical-devices/vitro-diagnostics/list-cleared-or-approved-companion-diagnostic-devices-vitro-and-imaging-tools. Accessed on: November 16, 2020.

 

Share This Post

Leave a Reply