A research team from Italy built their own AI algorithm with Aiforia® Create, Aiforia’s versatile AI development tool, for assisting in the process of kidney transplantation. We interviewed Dr. L’Imperio and Prof. Pagni to hear more about their journey so far.
Vincenzo L’Imperio is a Pathologist and an Assistant Professor of Pathology at the University of Milano-Bicocca. He works as a Practitioner Pathologist and Renal Pathologist in the Pathology Department of the Fondazione IRCCS San Gerardo dei Tintori in Monza. He is also on the board of the European Society of Digital and Integrative Pathology (ESDIP).
“We hope to give a proof of concept from the Galileo experience to other colleagues and confirm that this is the moment to start the collaboration between the human eye and AI for renal pathology diagnosis,” Prof. Pagni comments.
“One of the most important topics in renal pathology regards the pre-implantation kidney biopsy, which is usually assessed to give an overview of the possible donor health and the possible use of the kidney,” Prof. Pagni explains.
“The evaluation of the percentage of glomerular sclerosis, the amount of interstitial fibrosis and tubular atrophy, and arterial/arteriolar sclerotic narrowing are the most critical parameters, which are usually evaluated in the biopsies to give nephrologists a final overview of the health of the donor. The most used system is the so-called Remuzzi-Karpinski score, which combines the histological factors with other additional changes, assessing ischemia-related damage of the renal allograft.”
Prof. Pagni then explains that a tremendous clinical issue concerns the possibility of a general pathologist evaluating the Remuzzi-Karpinski score in their routine practice. There are challenges in this kind of evaluation by night or during a consultation on frozen sections for a pathologist who does not have dedicated training in renal pathology. The research team started with the Galileo project to solve this issue by giving the general pathologists a tool to help them in the score evaluation.
The team chose to call the project “Galileo” to honor the father of the scientific approach, one with roots in Italy. They also talk about a so-called “Ferrari analogy,” which relates to the annual Italian Grand Prix in Monza. “I imagine the Galileo facility of Aiforia as a Ferrari team. That is the perfect car, but you need the best pilots. So, in our mind, our group of Italian renal pathologists were the perfect pilots to bring this Ferrari to the final goal,” Dr. L’Imperio explains.
“We started imagining the Galileo tool for assessing the biopsies, asking ourselves which of the structures that we assess in the kidneys are easily recognizable and scorable by pathology with a good reproducibility, but similarly, they can still be troublesome for our colleagues that are not experts in the evaluation of kidneys, such as the on-call pathologists,” Dr. L’Imperio answers.
He imagines that the tool will be helpful for on-call pathologists who do not deal with kidney structures in their routine work and for renal pathologists such as himself, who want to save time on manual work. “I would use it because it's even more rapid than my eyes or my brain counting the structures and assessing the damage.”
Italy is currently suffering from a vocation crisis in pathology: it is challenging to find and recruit young pathologists for the institutions. “Renal pathology is a particular field of our job with the need for a great number of cases to increase the professional skills of pathologists in detecting challenging glomerular diseases and peculiar histological findings. Artificial intelligence can give a possible solution to improve the routine daily practice,” Prof. Pagni comments.
He explains that AI can help achieve greater inter-pathologist agreement and reproducibility in the evaluation of parameters and histological findings. Moreover, AI can reduce the number of tedious tasks and free up the pathologist’s time for other things. This is especially important when the resources are scarce.
The team has used the Aiforia® Platform for almost a year now and has found it perfect for pathologists with a clear clinical need and research questions but without a previous background in computer science. Dr. L’Imperio also comments that the cloud-based platform has made it possible for them to collect cases and collaborate together in real-time, all around Italy. They have also been able to share their experiences on how to “feed the algorithm” from different perspectives of specialist renal pathologists.
“Regarding technicalities of AI development, the presence of constant assistance by the Aiforia team has been another great added value of using this kind of a platform.” – Vincenzo L’Imperio, Pathologist and Assistant Professor at the University of Milano-Bicocca
Going forward, the team imagines that the Galileo tool will be beneficial for educational purposes, creating important teaching material for physicians in applying robust and reproducible criteria for the evaluation of specific kidney donor-related parameters.
“This is a great opportunity to build an archive of AI-guided slides to improve our knowledge in histology and in renal pathology. The partnership between the researcher of a public university and the company is very important to fill the gap between what we are learning in the academia and what the needs are from the real-world companies,” Prof. Pagni summarizes.
The team already has ideas for future projects, but they are keeping their cards close to their chest for the time being. “From a research point of view, we are really satisfied, and we are already thinking and brainstorming on what is the next use case to submit to the Aiforia team to develop together. I don't want to spoil anything, but we have another couple of ideas on the table,” Dr. L’Imperio hints.
One topic that we already know of, besides the work on kidney disorders, is related to HER2 low and ultralow in breast cancer, a trending topic in pathology. "At the moment, we are evaluating the partnership with the Aiforia team for the introduction of an AI-based tool for the evaluation of HER2 low cases in the breast histological section of breast carcinoma. This is a new tool that can help us in the challenging situation, and we hope that the final result can give the pathology community a new key for the solution of challenging routine cases for the definition of a specific and very important predictive biomarker."
Watch Dr. L’Imperio and Prof. Pagni’s video interview below: