Pro Medicus - AI Accelerator
PME signs collaboration agreement with UCSF for AI development
Dr Malte Westerhoff CTO and Visage Co-Founder “We have optimised Visage AI to enable both our own and 3rd party algorithms seamlessly in the clinicians desktop”
PME announced today a multi year research agreement with UCSF to collaborate on developing AI products with a view to commercialisation. My notes on this topic are outlined below.
Why will AI empower PME
Firstly PME have the platform - Most radiology AI highlights something from an image, and this requires a second popup window on a separate monitor or application, whereas Visage can overlay the AI on the diagnostic image in the same application - we all know that reducing friction increases usage. Being able to integrate is critical.
Secondly all Visage imaging is streamed using GPUs and most AI is run on GPUs not CPUs so all Visage customers already have the hardware in situ and ready for deployment without the need for any upgrades or installs.
How will AI likely be used?
PME have said for some time now that Radiology AI will evolve largely in 4 categories:
1- Triage, Prioritising a workload and ensuring that critical cases are seen first to save lives in a crowded environment.
2- Screening, being able to screen for early signs of things across a population
3- Aid to Diagnosis, Double checking what a radiologist looks at.
4- Automated Diagnosis.
How will AI commercialise in radiology?
PME have mentioned previously that Radiology AI will evolve in three buckets:
1- Internally developed FDA cleared AI products, PME currently have Breast Density in commercial use and are working on Breast Cancer, Bone Mineral Density assessment, Aortic Calcification and Body composition screening tools. All these are native to the Visage platform:
2- Collaboration with best institutions in the world to co-develop new applications. Like the one announced today with UCSF.
3- 3rd Party integrations to the Visage platform like the Elucid investment providing a partnership to integrate the FFR-CT and Plaque characterisation AI into Visage.
Some AI will be charged per click and some will be folded into price increases. It is unclear how the mix of these three categories will evolve but efficiency is the main driver, having AI easily accessible and frictionless will drive adoption, Radiologists don’t want 100 different applications to deal with.