Phedias Diamandis
• Neuropathologist, Dept. of Pathology, University Health Network
• Assistant Professor, Dept. of Laboratory Medicine and Pathobiology, U of T
• Scientist, Princess Margaret Cancer Centre
Professor Phedias Diamandis is the Principal Investigator with the Diamandis Lab II. His lab uses innovations in tissue engineering, chemical biology, deep learning and mass spectrometry-based proteomics to provide new insight and therapeutic approaches to different neurological disorders. We caught up with him to learn more about his work and where he sees Artificial Intelligence heading in the future.
What exciting projects are you currently working on?
I'm excited about a number of our current AI initiatives. We are currently focusing on using machine learning to automate histopathological image analysis and aid in the interpretation of the brain's electrical patterns and how they change with disease. We are still very early in translating AI to healthcare, so sometimes it's exciting to watch how our algorithms learn and interpret healthcare data across these projects. Sometimes they make human-like decisions and sometimes they make really odd, yet clever, decisions. Seeing both approaches and trying to piece together these observations with my clinical experience and training is one of the most interesting parts about working with AI.
What excites you the most about the possibilities of AI in healthcare?
Good healthcare is really expensive to provide and challenging to scale. It requires a lot of a society's resources and even with economies of scale, smaller cities and emerging economies have a difficult time providing adequate healthcare for all. AI helps digitize and automate many specialized manual tasks, such as medical image analysis, and could allow them to be rapidly disseminated at lower costs. I think this alone, makes AI a transformative and exciting technology for healthcare. Hopefully in the future it will provide tools that not only help mimic our current practices but also improve our ability to provide better care beyond automation.
What do you see as the biggest challenge to AI in healthcare right now?
While many aspects of healthcare can be broken down into evidence-based and data-driven decisions, other key components are inherently (and thankfully) subjective. These subjective components are much harder to quantify and automate. For example, how a patient chooses to balance potential benefits and side-effects of a particular treatment is inherently subjective and may change. This lack of clear "goals" makes training AI to deliver "personalized" healthcare decisions difficult. Finding a balance between these "scientific" and "humane" solutions to good decision making in healthcare is essential to maintaining our identity and dignity. We will need a lot of investment into the ethics and philosophy of defining "desired outcomes of care" before we are able to design AI to help in the most challenging aspects of medicine and healthcare.
What do you like to do when you aren’t working?
I like to play and watch tennis. It's been really fun growing up during the "golden era" of tennis and it is something I have enjoyed with my parents, spouse, friends and hopefully, soon, my own children! I learned so much about life from this game.
What advice would you give to STEM students considering an AI in healthcare career?
We need people with so-called "intellectual range" who can think like a healthcare provider but also computationally. I recommend that my students develop core competencies where they really excel, but also read a variety of things outside of their field so they can connect diverse ideas to ask and address clinically meaningful questions in novel ways. This is the hardest part of working with translational aspects of AI. I believe people with these types of skills will be the innovators in the years and decades to come.
