pathkids logo top logo
  Home
  Training days
  Cases
  Teaching
  Links
  Exams
_EQA
  Fun
Contact

THE DIGITAL PATHOLOGY REVOLUTION!

This is a draft of my article for the Royal College of Surgeon's bulletin on digital pathology.

The final published article is available at this link

Digital Pathology

 

What is pathology?

When most people think of pathology, Silent Witness or Dexter usually springs to mind but pathology is actually the umbrella term for all the laboratory and clinical specialities which analyse samples and specimens from patients; this includes microbiology, haematology, clinical biochemistry, immunology and cellular pathology.  Cellular pathologists (also known as histopathologists) such as myself, do carry out postmortems, but spend most of our time looking at glass slides of patient’s biopsies and surgical specimens under our microscopes.  In other pathology specialities, namely blood sciences, much of the testing process is automated and a numerical result can be generated which can be interpreted by most health care professionals (and some patients). In contrast, specimens in my department currently go through a highly labour intensive, mainly manual process getting the three dimensional tissue dissected and transferred onto one or many two dimensional stained sections mounted on glass slides. Once the slides are stained and ready, a pathologist interprets the changes in clinical context before producing a text or verbal report to the surgeon or physician looking after the patient. Since Virchow in the 1800s, this basic process has been the bedrock of tissue diagnosis and I’m not aware of a more specific test for cancer despite advances in imaging and ever more sophisticated innovations in genetics and molecular biology.   

 

What is digital pathology?

The glass slide and the tissue sample from which it is cut holds a tremendous amount of information about the patient and their illness, but unlike a numerical blood test result, very few people can make sense of the unique pattern on a patient’s slide, even those who know most about their patient, the managing clinician. Surgeons and physicians therefore rely on pathologists to interpret the microscopic images on a patient’s slide and provide useful diagnostic and prognostic information. Digital pathology was born (in its most simple form) out of necessity in some geographically isolated parts of the world allowing surgeons to get a tissue diagnosis at a hospital without a resident pathologist (1). Early “digital pathology tele-consultations” often involved intraoperative frozen sections being transmitted from a conventional microscope driven by a technician, to a pathologist hundreds of miles away in one of the big cities.  This basic form of “telemicroscopy” has more recently evolved into so-called “whole slide imaging” (WSI) which is where all the visual data on the entire microscope slide is scanned and stored digitally (2).   The latter (like radiology PACS) allows multiple users to view the slides simultaneously in different locations (see figure 1).



 

Figure 1. The diagram begins on the left with glass slide. At the top is the conventional method of a pathologist receiving the slide(s) who then often decides to send the case away if it requires a second/specialist opinion. Glass slides and the paraffin block are usually sent in the post with the risk of tissue being lost and damaged. If the case is very difficult the second pathologist may decide to send it to another specialist with further delay and risk. Alternatively the lower panel shows the advantages when the slide(s) are scanned and can then be viewed locally and/or remotely by any number of specialists, potentially at the same time to allow consensus opinion and diagnosis especially when the pathology can be simultaneously viewed with other clinical and radiological data.

An evolving solution for an emerging problem

There is a national and global shortage of pathologists due to poor  recruitment in recent years, combined with an increase in the numbers and complexity of diagnostic material we receive. There has been a considerable increase in the number of health professionals performing biopsies and carrying out surgical procedures without a corresponding increase in the number of pathologists (or laboratory resources).  Centralisation of most cancer services now means that suspected and confirmed cancer diagnostic biopsies need to be reported both at a peripheral (spoke) hospital, as well as at the specialist (hub) centre.  These slides have to be transported frequently between hospitals, leading to delays and a considerable burden of additional administrative work and consultant time.  Digitisation of pathology slides offers the potential of hub and spoke hospitals being able to share diagnostic information instantaneously.  Currently there needs to be a lead pathologist for every sub-specialty at every hospital, even for conditions which may only be seen at that centre a few times per year.   Increasingly specialised knowledge is required by all members of the diagnostic team when dealing with a more specialised surgical and medical workforce.  It makes more sense, and is in the patient’s interest therefore that a diagnostician who sees a lot of this disease, look at this type of biopsy, wherever they happen to be working, and be able to issue a prompt report and opinion on the clinical management. Like radiology scans, there is often a pathology slide “reporting backlog” and there is no reason once slides are digitised that many cases can be reported on the other side of the world if different time zones and costs prove beneficial.  

 

If the specialist pathologist is able to work remotely then there is also the potential of more flexible working.  An increasing percentage of our trainee recruits look to a career in pathology for its lifestyle benefits. Most of us work “office hours” and have no on-call commitments.  The potential of WSI and remote reporting offers a way to maximise the potential of a flexible family-friendly workforce. In addition, office space in hospitals and busy labs is at a premium and remote working is another way to potentially continue to tap the wisdom of recently retired senior pathologists who wish to continue to do some diagnostic work and help train the next generation away from the “shop floor”.

 

Potential Efficiency and Cost Savings

Increasing emphasis is now being placed on future cost savings that will be enabled by digital pathology. Physical storage of glass slides is now becoming more expensive than long term secure digital storage (see figure 2), even considering that each slide can hold at least a gigabyte of data when scanned at high magnification. The other immediate advantage over physical storage is of rapid access to current and previous images.  All slides showing suspected or confirmed cancer need to be reviewed at a MDT (Multi-Disciplinary Team meeting) at least one more time after the initial diagnosis. Keeping track of these slides is a logistical nightmare, particularly when they have to be transported to a different hospital for one or more regional MDT.   Much of the administrative workload of a hospital lab is consumed “chasing” urgent slides buried in consultant offices and corridors. Most laboratories deliver (push) multiple cases in a batch to each consultant without necessarily knowing whether that consultant is available and able to immediately report a given case. Backlogs of unreported cases are commonplace with the chronic staff shortages but importantly, clinically urgent work is often hidden amongst less crucial biopsies (see figure 3). WSI allows the digital workload to be efficiently categorised and arranged in terms of clinical urgency (or indeed by any other category). Importantly the pathologist is then able to “pull” cases individually from the triage screen whenever and wherever he/she is available to report a case, increasing efficiency and enabling early identification of a dynamic reporting backlog that can be promptly outsourced when necessary.

 

Figure 2. Rapidly decreasing cost of digital storage (source = http://honesthypocrite.blogspot.com/)

Figure 3. A typical UK pathologist’s office. Much of the delay associated with tissue diagnosis is associated with urgent cases being buried under other piles of routine work in offices and corridors due to the difficulties associated with keeping track of a (currently in most hospitals) purely analog, glass slide-based workflow.

Ergonomic and Visual advantages

Using a microscope for hours on end is a health hazard and many pathologists develop serious musculoskeletal problems. Viewing slides on a screen allows a more relaxed neck and back position and eliminates the current situation where a pathologist continually alters their neck position and visual focus from the microscope to a screen. Digital slides can be manipulated using a standard computer mouse but the greatest benefit and ergonomic experience is gained by the use of a “3d mouse” which enables the pathologist to move around in three dimensions as if flying a helicopter around the microscopic landscape. Similar to current radiology systems, digital pathology is best viewed on large flat screen monitors where current and previous slides are easily viewed side by side for easy comparison. Three dimensional reconstructions are also possible to enable the pathologist to better appreciate and understand the microanatomical details of the biopsy or specimen of interest and relate any sampled tumour to surgical margins for example.

Automated diagnostics, prognostics and predictives

It is my strong belief that automation of tasks currently carried out by human brains will truly revolutionise cellular pathology in the realm of cancer diagnosis and become pivotal in the new era of personalised medicine.  The human brain is very good at recognising patterns (it was clearly more important to our ancestors to correctly spot a lion than count the number of potential predators!), but quantification, measuring and counting is much more reliably carried out by computers. Every cancer has its own grading system, which aims to predict aggressiveness but most systems with more than two grades show only reasonable or poor reproducibility between different individuals, even when limited to those pathologists considered experts in their field (3).  Most grading parameters (for example severity of nuclear atypia) are subjective and some prognostic indices requiring counting hundreds of cells, and calculating mitotic ratios are very laborious and time consuming. Pathologists are aware that in most cases these systems do not critically guide management and generalists may rarely use certain grading systems, hence reliability is questionable in routine practice.

 

Increasing availability of biologic targeted oncological drugs requires testing of tissue sections, commonly with antibodies against an overexpressed oncoprotein receptor.  These “predictive tests” critically guide whether or not these expensive drugs are prescribed and predict whether the patient is likely to respond to the treatment.  The cutoffs between a positive and negative test, however is another subjective assessment liable to well documented variations between different laboratories and pathologists (4).  I have ethical concerns that whether or not a patient receives a potentially life giving drug like herceptin for example depends on my subjective assessment of staining intensity on the patient’s tumour cells, particularly now that I know there are computer algorithms which can be standardised to do the job for me [(5)and figure 4].

 

Figure 4. Computer assisted diagnosis is still in its infancy but there are already reliable digital algorithms available for removing subjectivity in the assessment of tumour cell proliferation rate, hormone receptor status, and predictive testing such as HER2 immunohistochemistry.

 

Many readers will be aware of the recent press releases about pigeons being able to diagnose breast cancer on digital histopathology slides with high (more than 90%) accuracy when “trained” by food reinforcement [(6) and figure 5]. Even more recently and remarkably, “machine-learning algorithms” have been used to successfully predict survival of lung cancer patients based on automated selection of histological features on their digital slides (7). Although pathologists will not yet be worrying about losing their jobs to avian colleagues or Marvin the robot, it does suggest that in the not too distant future, histological patterns of malignancy could be detected and classified by non-human neural networks. While unlikely to replace the job of a pathologist, these automated processes are likely to work in parallel with the pathologist to help improve the accuracy, reliability and efficiency of histological testing.

 

Figure 5.  Bird brains or brainy birds? Pigeons have recently been shown to be able to reliably recognise  novel histological images of malignant tumours after being trained with food reinforcement using a series of “good” examples of  benign and malignant breast tissue slides.

 

Promoting the speciality for the next generation

I have already alluded to the looming recruitment crisis in pathology and yet the demand for increasing numbers of detailed and specialised test results from our workforce continues.  Radiology faces similarly increasing demand but the digitisation of scans has allowed more flexible working, rapid outsourcing and other benefits already described. In addition, digitisation has made radiology more clinically conspicuous enabling healthcare professionals to view scans as soon as they are available in theatre, wards, clinics and primary care. This has undoubtedly made the speciality more appealing to medical students and junior doctors who can appreciate its ubiquitous clinical relevance and importance.  Digitisation of slides would allow pathology to be similarly ubiquitous, particularly when combined with computer algorithms to give the clinician meaningful and more objective information to guide management. Considering pathology speciality training, there are a huge range of diseases that trainees must learn about and learn to recognise during their now shortened training period. WSI enables archiving of real cases of rare and unusual diseases which can subsequently be at their fingertips within minutes. Currently as has already been mentioned older slides are more difficult to retrieve and are usually faded and unrecognisable when pulled from off-site storage. Trainees usually have to travel to a different city or region to view slides of certain subspecialities such as neuropathology, sarcoma or paediatrics which cannot be seen at a DGH or non-specialist centre. Digitisation will undoubtedly raise the profile of the speciality, and combined with the potential to work flexibly with state of the art technology rather than in the hospital basement with dusty glass slides, might even change the opinion of many that histopathology is the last respite of the socially inept (8).

 

Figure 6. The digital pathology workstation incorporating automated grading systems and predictive testing algorithms may make pathology more appealing to a new generation of IT-savvy medical students and junior doctors.

Problems to overcome

The primary problem with the widespread implementation of digital pathology in the NHS is currently the cost. Slide scanners, workstations and improved IT equipment are all “on costs” since glass slides still need to be produced prior to scanning. The NHS is notoriously broke and pathology seems unlikely to be a top funding priority in the next few years. Pathology departments need to look carefully at potential efficiency savings of WSI to convince funding organisations that they can receive a significant return on their investment.  Long term secure storage costs form a vital part of this calculation, although as I have already mentioned, digital storage is rapidly becoming significantly cheaper than physical storage of glass slides and this fact should form an important driving force in full adoption as the cost of digital storage continues to fall.  Network speed is also rapidly increasing and this will need to be optimised. WSI files are considerably larger than radiology images but pathologists usually look at and interpret most areas of a slide in a matter of seconds. A balance has to be struck between storing all the visual information from the slide at high resolution, and enabling rapid access to local and remote users. Rapidly improving pyramidal imaging processing algorithms (as used in Google Earth) allow “high altitude” views to be quickly loaded from the network making primary diagnosis more feasible at current NHS network speeds (see figure 7). Vendors and IT services will also need to ensure that regional, national and international firewalls will not preclude easy sharing of cases between experts so that these time-saving advantages can be used to benefit the patient.

 

C:\Users\braceyt\AppData\Local\Temp\Presentation1.jpg

Figure 7. The resolution of a WSI at high magnification is roughly equivalent to an image of a sports stadium detailed enough to visualize a dewdrop on an individual blade of grass (inset left). Pyramidal image processing (inset right) enables the low power “thumbnail” to be loaded quickly to allow the pathologist to zoom into the tile of interest without having to load the entire high resolution baseline image all at once.

 

There are potential legal issues with the use of any new technology which may potentially impact patient care, particularly if remote reporting spans across different regions and even worldwide.  The UK Royal College of Pathologists have recognised the importance of digital teleconferencing for second opinions (9), but are yet to release clear guidance on the use of WSI for primary diagnosis and is still prohibited for this purpose by the United States FDA. There are now however UK (10) and international (11) validation studies showing “non-inferiority” of WSI in the clinical arena which are the first step towards more widespread recognition and acceptance of digital pathology.

Finally one of the biggest problems to overcome will be to persuade notoriously conservative pathologists to change and take advantage of this new technology.  The rapid and efficient distribution of work may impress many, but I think it will be the automation of laborious subjective tasks like tumour grading which will eventually convince even the most resistant to change, and more importantly enable surgeons to get the answers they need before their patients’ wounds heal.

 

References

 

1. Weinstein RS, Bloom KJ, Rozek LS. Telepathology. Long-distance diagnosis. Am J Clin Pathol. 1989;91(4 SUPPL. 1).

2. Pantanowitz L, Valenstein PN, Evans AJ, Kaplan KJ, Pfeifer JD, Wilbur DC, et al. Review of the current state of whole slide imaging in pathology. J Pathol Inform. 2011;2(around 1999):36.

3. Delahunt B, Egevad L, Samaratunga H, Martignoni G, Nacey JN, Srigley JR. Gleason and Fuhrman no longer make the grade. Histopathology. 2016. p. 475–81.

4. Sheffield BS, Garratt J, Kalloger SE, Li-Chang HH, Torlakovic EE, Gilks CB, et al. HER2/neu testing in gastric cancer by immunohistochemistry: Assessment of interlaboratory variation. Arch Pathol Lab Med. 2014;138(11):1495–502.

5. Brügmann A, Eld M, Lelkaitis G, Nielsen S, Grunkin M, Hansen JD, et al. Digital image analysis of membrane connectivity is a robust measure of HER2 immunostains. Breast Cancer Res Treat. 2012;132(1):41–9.

6. Levenson RM, Krupinski EA, Navarro VM, Wasserman EA. Pigeons (Columba livia) as trainable observers of pathology and radiology breast cancer images. PLoS One. 2015;10(11):1–21.

7. Yu K-H, Zhang C, Berry GJ, Altman RB, Ré C, Rubin DL, et al. Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun [Internet]. 2016;7:12474. Available from: http://www.nature.com/doifinder/10.1038/ncomms12474%5Cnhttp://www.ncbi.nlm.nih.gov/pubmed/27527408%5Cnhttp://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=PMC4990706

8. Schubert M. The Last Respite of the Socially Inept? The Pathologist [Internet]. 2014 Dec;1(3):18–25. Available from: https://www.google.co.uk/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0ahUKEwjA9tTWmcnPAhUHJcAKHVk9CS8QFggeMAA&url=https%3A%2F%2Fthepathologist.com%2Ffileadmin%2Fpdf%2FTP_0314_Issue.pdf&usg=AFQjCNE9AyStktAXZYraXEP9-YOzzUv71Q&sig2=ss79yHfRIF

9. Lowe J. Telepathology : Guidance from The Royal College of Pathologists October 2013. 2014;(October 2013):1–11.

10. Snead DRJ, Tsang YW, Meskiri A, Kimani PK, Crossman R, Rajpoot NM, et al. Validation of digital pathology imaging for primary histopathological diagnosis. Histopathology. 2016;68(7):1063–72.

11. Cheng CL, Azhar R, Sng SHA, Chua YQ, Hwang JSG, Chin JPF, et al. Enabling digital pathology in the diagnostic setting: navigating through the implementation journey in an academic medical centre. J Clin Pathol [Internet]. 2016;jclinpath-2015-203600. Available from: http://jcp.bmj.com/lookup/doi/10.1136/jclinpath-2015-203600

The scheme above was borrowed from a webinar Leica recently broadcasted. It is still available to view at this linkIt was a fascinating lecture about what I feel has to be the future of our speciality. All aspects of telepathology were covered; slide scanning, archiving, live videoconferencing, quality and safety aspects, image analysis, cost, time and manpower savings.

Dr Tim Bracey

Consultant Pathologist

Derriford Hospital Plymouth

 

Pathkids recommended books. Use the following search box to find Amazon books directly from Pathkids