Trans-Continental Pathological Philanthropy

The Agha Khan University Hospital (AKUH), based in Nairobi, Kenya, is a private, not-for-profit hospital that falls under the auspices of the Agha Khan Foundation. When Shahin Sayed was appointed as Medical Director of the Department of Pathology and Laboratory Medicine at AKUH, she had ambitions of elevating the department to become a referral center for biomarkers for eastern and central Africa. Such a vision would require some bold and innovative thinking – and perhaps a strategic partner…

Enter Diagnexia, which provides artificial intelligence (AI)-empowered pathology resources for second opinions and diagnoses. With their network of pathologists and expertise in both digital pathology and AI solutions, Diagnexia was well placed to provide a timely boost in pathology power to AKUH. 

We spoke with Runjan Chetty, Chief Medical Officer at Diagnexia, to find out more about the resulting partnership with AKUH – and what it will mean to both parties.

Could you briefly introduce the partnership that you forged and what it aims to achieve?
 

It was the most serendipitous meeting of minds. When AKUH first contacted me, I was at a point in my career where I was expanding my work from academic pathology to diagnostics in the private sector. This presented the opportunity to reach out to low- and middle-income countries, which are under-resourced, pathology- and technology-wise, to try and provide access. 

The conversations led to how AKUH and Diagnexia could work together for mutual benefit. Eventually, a partnership was founded on us creating and sharing a digital pathology network with the AKUH. They acquired and validated a glass slide scanner, able to generate whole slide images; we were able to add a secure portal for upload and exchange of pathology cases.

This system facilitated the transmission and exchange of cases in a matter of hours and having second opinions back within a day. That has been very beneficial to AKUH in terms of improving turnaround times and facilitating access to that critical second opinion, where required, so that patients’ treatments can ensue as soon as possible.

In what other ways will this digital pathology network help AKUH enhance its patient care?
 

Very few, if any, pathology labs in Kenya are adequately staffed to manage the population. I think there is a general shortage of pathologists there that’s reflective of what's going on around the world.

We are trying to fill that void in a small way by providing access to more pathologists. And the partnership goes beyond second opinions; we can also help with external quality assurance. AKUH pathologists can select cases randomly or in a more directed fashion, and send them to the network for our subspecialty to vet the quality of the report that is contained within. Diagnexia’s pathologists can look at some of the slides of the case and then write a review back, which is important for AKUH’s accreditation purposes. Accrediting bodies, such as the College of American Pathologists, require a percentage of cases in any department to have external peer review, and we are providing that for them. 

Alongside that, our network consists of some key leaders in the various subspecialties who are very happy to provide mentorship to the local pathologists, along with help with publications. We can partner, say, the local kidney pathologist with a kidney pathologist in the UK, they exchange information and questions, and mentorship follows. 

We’re also very keen to foster research opportunities. Case material and pathology that's seen in Kenya is different to the rest of the world, and we can enhance that. And AKUH pathologists have the ability to enrich our cases in the Western world. I think that's very important. 

We have several research projects running on various levels, from purely descriptive clinical studies to more in-depth molecular studies, where pathologists in our network are interacting with local AKUH pathologists. The scheme is in its fledgeling state but certainly has great potential. 

And, in a truly collaborative fashion, we also want to develop AI models and tools with AKUH, using their material for deployment locally as well. It’s a multifaceted arrangement, and very bilateral, not just a unilateral one. AKUH’s input is valuable.

How might AI tools be beneficial for AKUH?
 

Perhaps I could preface my answer by addressing the great debate that’s out there in the pathology world: will AI replace pathologists? 

And the answer is no. Clearly, you need nuanced human knowledge to put together the pathological findings in context with the clinical ones, and to communicate the conclusions with medical colleagues who are going to treat the patient. We simply can’t have a standalone AI machine, for want of a better word, producing pathology results. 

However, having these computer-aided diagnostics tools can take away much of the subjectivity that besets pathology. For example, if we look at the subtle differences between a very reactive epithelium and a low grade dysplastic epithelium, there are gray areas where the discriminatory power of AI can aid the pathologist to say, “No, the tool concludes that it’s not dysplastic.” 

There are several applications of AI that can augment the pathologist’s practice, boost efficiency, and free us up to do other tasks that perhaps will contribute more to patient management. In terms of screening tools, we can take the example of a biopsy for inflammation in the stomach. Gastritis is a very common complaint, for which H. pylori is the offending organism in the vast majority of cases. Typically, we spend hours and hours hunting for these organisms so that the appropriate treatment can be instituted – histological evaluation of the gastric biopsy is the gold standard for diagnosis. This works well where the organisms are plentiful, but less well if, say, the patient has been pretreated and presents with a diminishing number of organisms. We're working on an AI tool to identify biopsies that have a proclivity to contain H. pylori organisms. It takes into account multiple parameters, flags likely infection, and orders the relevant immunohistochemical stain, which then comes to the pathologist, saving time. If that didn't happen, then the pathologist would have to screen the slide for signs indicative of H. pylori infection, then go back and order the special stain, and so on. There’s an efficiency gain purely in that one scenario. 

We’ve also been looking at tools that will triage cases. People over certain age bands are recommended to have a routine colonoscopy in the interests of preventative bowel cancer screening. AI tools are available that will triage the screening test results, and stratify them as high, medium, or low risk for cancer, moving high-risk results to the top of the pathologist’s list for priority review. These programs increase early detection rates, hence morbidity is decreased. 

Then there are tools that help with counting and quantification – another time-consuming and somewhat subjective exercise for pathologists to perform manually. Tasks such as mitotic counting can be done in an automated fashion. Again, this is performed under human control, because the pathologist knows where to look. But the AI assistant performs the counting, making reporting much quicker and more efficient. 

And, finally – something that has come into vogue in the last four or five years in cancer diagnostics – AI tools can help interpret biomarker profiles, which are important for informing treatment decisions. Again, counting viable tumor cells, such as PD-L1-positive cells, on a slide can be a very laborious, drawn out process. This is the sort of terrain in which AI can save time as well as produce very objective key results. And you can rely on them being reproducible. The patient undoubtedly benefits from that as well. 

Those are some of the scenarios for which we're going to collaborate with the AKUH. Then, once the tools are developed, they can roll it out in their own practice. Not only will this help AKUH in its ambitions to become a referral center for biomarker testing, it will also provide added value to patient care.

Are the resources in AKUH sufficient to support digital pathology and AI?
 

We have tried to minimize the reliance on local data storage in AKUH as much as possible. The mode of operation is for AKUH to scan and then upload images to a shared drive, from which it is uploaded into Diagnexia’s cloud storage for access by our pathologists. From our side, we have plenty of data-protection-compliant cloud space. And from AKUH’s side, once the drive is full, the images can be removed from their servers once the case has been reported and they simply retain the glass slides. 

When AKUH moves to a fully digital workflow, of course, then internal solutions will have to be found for data storage. But, at this stage in this working relationship, there isn’t a huge demand on data storage for our partner.

Why are global health initiatives important to you personally – and to Diagnexia as an organization?
 

We want to contribute to pathology not only as a commercial exercise, but also with a philanthropic outlook. We offer the second opinion service to our partner pro bono. The participating pathologists are very keen and happy to take on these cases without any personal charge. There's a small handling cost, to compensate those who are doing the background work, but it is minimal. 

That ethos is very important to us as a company and to me as an individual. The shortage of pathologists and resources worldwide is quite alarming, and one of our goals was to try and provide global access. We want to encourage and foster these types of partnerships in terms of trying to improve health care via the foundational pillar of pathology – good pathology leads to good patient outcomes. We can utilize the expertise in our global network of more than 200 subspecialty experts who are willing to give their time and also endorse the philanthropic aspect of what we do. It’s a really positive thing for us. 

What other philanthropic activities is Diagnexia involved in?
 

I’ve reached a stage in my career where teaching and developing the new generation of pathologists is very important to me. So we also supplement our partnership activities with an educational program whereby we present a challenging teaching case every month. We’re also starting a new webinar series. For example, an expert liver pathologist might provide a demonstration of a liver biopsy. People from around the world can listen in for that learning, incorporate it into their safe practice, and produce better diagnoses as a result. 

And, finally, we run an annual symposium in Oxford, UK. Last year, we welcomed delegates from 20 countries. We take the first 25 trainees who register free of charge, irrespective of where they come from. All they have to do is travel here, and we take care of the accommodation, the food, and other costs. We do that because we want to encourage those early career professionals to come and benefit from exposure to these international experts who come in to lecture. We provide the opportunity to meet with world experts, face to face – maybe have a cup of tea with them. They have these role models at their disposal, so to speak, and we find it has an immense effect on their future careers and their development. We also have demonstrations of AI and digital platforms so our delegates can see the benefits of a digital workflow, which is really beginning to gain traction – particularly here in the UK.

Are you looking at any other global health partnerships at the moment?
 

Absolutely. We are in discussions with several potential partners in various global locations. We are focused on globalization of pathology!