CompuMed and its team of highly experienced physicians provide specialized diagnostic interpretations to their clients who need answers in high urgency situations. Organ Procurement Organizations (OPOs) which provide organs and tissue for transplant centers need critical information relating to for example pathology, cardiology, or radiology to improve donor management, decrease donor evaluation time prior to organ recovery, and to decrease the risk of organs being declined by transplant centers.
Many of the specialized physician interpretations needed for organ management and evaluation are best done by those that have expertise in that specific area. In addition, these services are often needed in a fast turnaround time, often at night, over the weekends and holidays. Many of the donor hospitals are often not very large centers, and even in larger hospitals, key resources tend to be mainly allocated to treating living patients.
“The OPO must often make do with late, and/or nontransplant-specific interpretations. Due to the large number of donor hospitals, over 50% of the donors come from a hospital that has less than two donors a year, this process involves a large number of pathologists, which by its very nature, results in inconsistencies,” explains Carie Kadric', Vice President of Clinical Operations at CompuMed.
Part of the evaluation process by pathologists for transplantation may include a biopsy performed at the time of procurement. The transplant center must view the biopsy by viewing the actual slide, therefore it must be physically transported, meaning that the initial hospital must provide a “provisional acceptance”: they will accept the organ and once they receive delivery of the organ and slide, they can make a final decision. This does not allow multiple transplant centers to view the slide in a timely manner and adds hours of time to an already procured organ, increasing the probability of it degrading.
The ability to now digitize slides allows a smaller more specialized reading group of pathologists to perform slide interpretation creating more consistency by reducing the possibility of inter-observer bias. Digitized slides also allow multiple transplant centers to be able to view the slide remotely while reducing the trauma on the organ from multiple biopsies.
Increasing the speed and precision of transplant pathology with AI
CompuMed wanted to investigate the benefits of using AI for the pathology process of their services, focusing on two specific functions. Firstly, in improving the “screening” of the biopsy slide prior to it reaching the remote pathologist. There are a number of thresholds that are required to be met for a biopsy to be an acceptable representation of the organ. For example, a minimum number of Glomeruli are required to be visible in a kidney biopsy, and the evaluation of the quality of the digital slide and the viable biopsy area are also critical.
“We felt that with the help of artificial intelligence we could greatly increase the confidence that what was being sent digitally to the pathologist would be acceptable and readable. This was especially important if the technician making the slide was on call as they may have already left by the time the pathologist indicated the slide or biopsy was unacceptable and they needed it to be redone,” describes Carie Kadric’.
The second function was to assist the pathologist in their evaluation to help improve efficiency, accuracy, and consistency. A renal slide from a kidney biopsy may take 10 to 15 minutes to read manually. Much of the time is spent covering the complete slide to count Glomeruli and determining if any are sclerosed, or damaged. With the assistance of AI CompuMed anticipates they will be able to increase the speed and accuracy of this evaluation.
“We wanted to partner with a company that not only had a full AI platform that we could use ourselves, but we also wanted the flexibility of utilizing a platform that had a team that could do the AI development for us. In this way, we could utilize the AI team in the initial stages and any time we did not have the bandwidth,” explains Carie Kadric’. “ When we evaluated platforms, Aiforia came up at the top. But what really made the difference was the ease of working with their team. We feel the project has exceeded our expectations, not only with the resultant product, but also the value of the deliverable versus the cost.”
CompuMed is currently using Aiforia’s cloud-based platform for kidney transplant pathology and looking to expand to other organs such as the liver. The kidney project is using deep learning AI through Aiforia to count Glomeruli and to indicate sclerotic ones, giving pathologists confidence in their decision making by providing them with fast, high-precision data available to them and their colleagues remotely. Quality control (QC) means are also being studied with Aiforia’s AI platform, as CompuMed aims to measure each pathologist’s estimations to the ones produced by the AI Models; comparing them across other pathologists to identify potential areas of improvement.