February 2018: Hands-on AI

There is a lot of talk about how artificial intelligence will affect radiology, but not much about the AI systems already in use. It is therefore timely that Radboud University and EuSoMII are organising a Hands-on AI Day in Nijmegen on April 21, 2018.

Nine AI-products, including BoneXpert, are selected for presentation and practical use.

Artificial Intelligence can be defined as a computer program that automates a cognitive task. It is implied that the task involves a considerable amount of complexity, so for instance a DEXA scanner which automatically determines BMD by measuring the X-ray attenuation, or a tool that semi-automatically measures the circumference of a child’s head in an ultrasound image, are probably too simple to be called AI.

AI describes what the product does, not how it was engineered, so the technology behind AI can be a rule-based expert system, a classical computer vision algorithm, a machine learning system, or a deep neural network.
The determination of bone age from a hand X-ray is a holistic interpretation of the shape and intensity patterns of each bone – it is not sufficient to use simple features like the relative width of the epiphysis. BoneXpert automates this interpretation using machine learning from thousands of X-ray examples and was launched as a medical device in 2009, and it was one of the first AI systems in use in radiology.

The hands-on day will give radiologists a unique opportunity to experience the impact of AI, and it will surely inspire to form an opinion on questions like

  • How does it feel to work with AI?
  • What are reasonable ways to validate AI systems?
  • What new roles of radiologists emerge from the adoption of AI?
  • How will AI affect education of radiologists?