AI-assisted healthcare is reaching maturity in many applications and could alleviate some of the capacity gap increasingly faced by health systems . Over the next three podcasts we focus on artificial intelligence tools designed to assist directly with clinical practice.
Most commonly reported on are the algorithms capable of pattern recognition on medical images, that in some settings perform as well or better than expert diagnosticians at classifying disease. AI models have also been developed to perform regression analyses more complex than classical risk stratification aids.
The standard statistical algorithms used to solve these problems struggle when many variables are introduced, in which case deep learning models that mimic brain networks are sometimes a powerful alternative. In this episode we explain how machine learning algorithms are trained on particular tasks and where there are risks of error and bias being introduced.
In part 2, we identify the ergonomic issues that affect practical implementation of AI tools in the clinic and in the decision cascade. And in the final episode of the series we discuss the questions that regulators and lawyers should be asking of this new technology and what role natural language processors might have in medicine.
Guest
Dr Ian Scott FRACP MHA MEd (Director of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital; Professor of Medicine, University of Queensland)
Production
Produced by Mic Cavazzini DPhil. Music licenced from Epidemic Sound includes ‘Thyone’ by Ben Elson, ‘Broke No More’ by Cushy, ‘Desert Hideout’ by Christopher Moe Ditlevesen and ‘Alienated’ by ELFL. Music courtesy of Free Music Archive includes ‘Capgras’ by Ben Carey. Image by Olemedia licenced from Getty Images.
Editorial feedback kindly provided by physicians; Rhiannon Mellor, David Arroyo, Aidan Tan, Joseph Lee, Rachel Murdoch, Michelle Chong, Phillipa Wormald and digital health academics; Paul Cooper and Natasa Lazarevic.
Key References
Demystifying machine learning: a primer for physicians [Scott, IMJ. 2021]
Clinician checklist for assessing suitability of machine learning applications in healthcare [Scott, BMJ Health Care Inform. 2021]
Please visit the Pomegranate Health web page for a transcript and supporting references. Login to MyCPD to record listening and reading as a prefilled learning activity. Subscribe to new episode email alerts or search for ‘Pomegranate Health’ in Apple Podcasts, Spotify, Castbox, or any podcasting app.
AI-assisted healthcare is reaching maturity in many applications and could alleviate some of the capacity gap increasingly faced by health systems . Over the next three podcasts we focus on artificial intelligence tools designed to assist directly with clinical practice.
Most commonly reported on are the algorithms capable of pattern recognition on medical images, that in some settings perform as well or better than expert diagnosticians at classifying disease. AI models have also been developed to perform regression analyses more complex than classical risk stratification aids.
The standard statistical algorithms used to solve these problems struggle when many variables are introduced, in which case deep learning models that mimic brain networks are sometimes a powerful alternative. In this episode we explain how machine learning algorithms are trained on particular tasks and where there are risks of error and bias being introduced.
In part 2, we identify the ergonomic issues that affect practical implementation of AI tools in the clinic and in the decision cascade. And in the final episode of the series we discuss the questions that regulators and lawyers should be asking of this new technology and what role natural language processors might have in medicine.
Guest
Dr Ian Scott FRACP MHA MEd (Director of Internal Medicine and Clinical Epidemiology, Princess Alexandra Hospital; Professor of Medicine, University of Queensland)
Production
Produced by Mic Cavazzini DPhil. Music licenced from Epidemic Sound includes ‘Thyone’ by Ben Elson, ‘Broke No More’ by Cushy, ‘Desert Hideout’ by Christopher Moe Ditlevesen and ‘Alienated’ by ELFL. Music courtesy of Free Music Archive includes ‘Capgras’ by Ben Carey. Image by Olemedia licenced from Getty Images.
Editorial feedback kindly provided by physicians; Rhiannon Mellor, David Arroyo, Aidan Tan, Joseph Lee, Rachel Murdoch, Michelle Chong, Phillipa Wormald and digital health academics; Paul Cooper and Natasa Lazarevic.
Key References
Demystifying machine learning: a primer for physicians [Scott, IMJ. 2021]
Clinician checklist for assessing suitability of machine learning applications in healthcare [Scott, BMJ Health Care Inform. 2021]
Please visit the Pomegranate Health web page for a transcript and supporting references. Login to MyCPD to record listening and reading as a prefilled learning activity. Subscribe to new episode email alerts or search for ‘Pomegranate Health’ in Apple Podcasts, Spotify, Castbox, or any podcasting app.