Earlier this month, AuntMinnie.com hosted a virtual conference during which a series of online lectures on advances in artificial intelligence (AI) were delivered by six experts. Although the presentations were scheduled at a time convenient for viewers in the USA and Europe, it is likely that many people were unable to take advantage of the opportunity to learn about this rapidly evolving field. The good news is that all these excellent lectures are now freely available and can be viewed at your leisure.

Bradley Erickson of the Mayo Clinic in Minnesota asked, “Deep learning in radiology: can we trust it?” He showed how the number of research articles on deep learning – also known as convolutional neural networks – has increased exponentially during the past five years, but argued that while AI is incredibly powerful, we may not always understand the subtle features and unintended consequences. Maryellen Giger of the University of Chicago considered the efforts to support the development of AI for diagnosis and treatment of Covid-19. Given that we are still dealing with the pandemic, this vital research is ongoing.

While the radiology community has been quick to embrace AI, that enthusiasm has apparently been dampened by a lack of reimbursement. Melissa Chen from the University of Texas (seen left) addressed this topic with a lecture entitled, “How to get paid using radiology AI,” and introduced a series of tools based on the Medicare payment model. Emily Conant of the University of Pennsylvania spoke on AI applications in mammographic screening, and she argued that, while AI has considerable potential, there are still some challenges that need to be addressed before widespread adoption.

Amy Patel of the University of Missouri in Kansas City presented her experience in using the AI tools from Koios Medical for hand-held ultrasound to detect breast cancer. Among her findings were fewer false positives, fewer false negatives, and a reduction in unnecessary biopsies. Importantly, the time saved in making a diagnosis was up to two minutes which led to greater patient throughput.

Peter Chang of the University of California, Irvine (seen right), who is both a neuroradiologist and software engineer, provided practical advice for evaluating more than 300 AI medical imaging tools now available in the marketplace. He highlighted various tools, including triage to improve workflow, image quality improvement, oncology applications and tools for segmentation and quantification. If you’re interested in AI applied to medical imaging, and you have a few hours to spare – perhaps on a rainy day – then these six lectures will provide a useful tutorial.