Though it has the potential to be truly revolutionary, the use of AI in pathology and other areas of medicine is ‘still in its infancy’ as was explained by collaborators and colleagues in a recent paper for the Journal of the American Society of Nephrology (JASN).

AI itself is still maturing – being developed and refined – and debates are ongoing about how it can be deployed ethically, particularly in fields like medical research and healthcare.

So, of course, there may be scepticism about the adoption of AI in something as serious as, say, pre-implantation biopsy analysis ahead of renal transplants. But, as we outlined in our JASN paper, this is one of the areas that has the most potential for radical enhancement – something that could save lives and, for the NHS, millions of pounds a year.

At present, the process of characterising chronic injury in donor kidneys is far from a perfect science. As we’ve shown through our groundbreaking work with the University of Cambridge’s Office for Translational Research (OTR), however, by using AI to automate the process of screening and scoring kidney biopsies, we can make it drastically more efficient.

This is achieved by more accurate assessments as to the suitability of organs for transplants, meaning less have to be rejected. To date many organs are rejected largely based on the age of the donor, as there naturally tends to be organ deterioration with age.

The bold vision of a national digital histopathology service could drive up the volume and standard of renal transplantation, marked by a reduced need for dialysis and a more efficient process of getting patients to transplants in the first place.

To reach the stage where clinicians not only trust but favour the use of this currently emerging technology, we must demonstrate its efficacy, efficiency and safety.

It is our duty, as AI technology visionaries, to earn that trust through diligence and the development of practical, workable solutions that improve accuracy and, ultimately, patient outcomes. We must demonstrate to practitioners that the AI models themselves are ethical and explainable, and continually being monitored and challenged to ensure the best results and avoid unintended biases.

This will require a large-scale effort with continued learning and listening on our part too. But already we have the very real promise of the much greater volumes and speeds of tissue processing that AI can offer over present systems. Already we know that certain decisions and judgements within the broad field of histopathology can be at best inefficient, and at worst impractical, when left to human assessment alone.

What our work in this nascent but evolving field shows is that where these human error-prone gaps exist, AI could potentially be a solution.

For a start, it could increase the volume and efficiency of biopsy slides or samples being tested. Or the science could be more far-reaching, and transform the way organs are assessed pre-implantation, beyond even what we are capable of envisioning today.

As I wrote for this publication in May, we all have a responsibility to use AI in an ethical way to ensure clinical soundness and patient safety.

If we, as technology providers, are going to earn the trust of the profession, we must be as assiduous in our efforts as we are scrupulous in our standards. We cannot expect medical professionals to accept any less. 

Simon Tilley is Global Lead for Healthcare and Life Science at SAS.
Go to sas.com