Introduction
As part of the Turing-Roche Partnership Community Scholar Scheme, the Translational Science Methods Club (T-SciM Club) aims to bring together members of Turing and Roche as well as the wider academic and clinical community to discuss different methods that help us translate from data to science to clinic.
About the event
The theme of this T-SciM event will be explainable AI (XAI). This refers to a range of approaches that seek to design or evaluate ways to make the outputs of artificial intelligence algorithms more interpretable. Such explainability is important to better understand the mechanisms underlying the models computations and determine how trustworthy and generalisable their results are.
We will be joined by 3 speakers, who will give short talks introducing their work on XAI, followed by a panel discussion with questions and prompts from the audience.
Dr Ahmed Salih's talk will introduce current approaches of evaluating XAI, weakness, limitations and how they are implemented in cardiology applications where the risks of misinterpreting results can be severe.
Sophie Martin will present her work on utilising amyloid-PET imaging and a 3D convolutional neural network to classify amyloid-positive and amyloid-negative individuals, which offers a unique opportunity to validate and assess whether the important brain regions according to the “black-box” model are biologically relevant.
Dr Tapabrata (Rohan) Chakraborty will present applications of transparent AI in healthcare through two distinct approaches – that of explainable models (with the example of hybridisation of data driven deep learning with interpretable mechanistic models) and trustworthy decisions (with the example of uncertainty quantification of individualised predictions by conformal analysis) and ask which is better for responsible and safe AI for personalised healthcare and precision medicine.
This event is specifically targeted to early career researchers, including those who have previously or intend to apply this framework in their own research as well as those simply interested in learning more and having a discussion about its potential applications in science.
Contact
As well as having a Q&A at the event, we also hope to have discussions about this theme pre and post the event via our Turing-Roche Slack Workspace which you can join here. Once you are a Slack member please join the channel #translational-science-methods.
If you are interested in being a speaker at one of our future events or want to give feedback, then please do get in touch with our T-SciM organiser Sarah Buehler at [email protected].
Watch now
You can watch a recording of this event here.
