One of AIMed’s signature events of the year – AIMed Europe 2019 embarked today (17 September). Near 300 delegates, speakers, and company representatives from around Europe and other parts of the World gathered at Shoreditch Town Hall in Central London for a two and a half
Although most attendees expressed they are attending the event for the very first time, this does not deter an active exchange of opinions and discussions around technologies and innovations. The interaction was further facilitated by AIMed’s effort to go paperless. A mobile application designed by the platform’s Head of Marketing not only digitized the event’s
On top of which, the relaunched version of AIMed magazine also made its debut. Known as AIM, the publication continues to be made available in print and download by subscribers online. The new issue showcased a special cover feature of the founding father of AI in medicine, Dr. Edward Shortliffe as well as a special hospital report, an overview of AI landscapes in Europe, columnists’ insights, research papers and book reviews, and many more.
Dr. Anthony Chang, Founder
The AI reality check
As more healthcare professionals get involved in developing AI-driven solutions, AIMed Europe is no longer about celebrating what new technologies can do, but pointing out what can be done better. Dr. Chang said in the opening session, machine learning (ML) may produce high accuracy in seeking out for abnormalities. However, this does not translate to a change in disease outcome or a change in patients’ behavior. For example, by telling a patient he or she has a higher chance of getting diabetes is not going to change how the patient behaves unless it is coupled with an executable intervention.
Likewise, even though data is now considered the new oil, in healthcare, data is still a kind of crude oil because of how unstructured and segregated they are. Researchers’ preference for deep learning over ML means they are more likely to fall into an AI Blackbox trap or the inability to explain why an AI-driven medical solution arrives at its conclusion and as a result, making it challenging to underline and undermine errors.
Dr. Matthieu Komorowski, Clinical Senior Lecturer, Imperial College London; Consultant, Intensive Care Unit, Charing Cross Hospital, and Visiting Scientist, Lab of Computational Physiology, Massachusetts Institute of Technology described the pitfalls of ML algorithms during his presentation. He believed most algorithms are not clinically useful. Those that are helpful may not be allowed to use on patients for one reason or the other. In certain cases, such as the sepsis prediction alert, are often turned off by medical professionals because of the high tendency of a false alarm.
Echoing Dr. Chang’s comment, Dr. Komorowski said having algorithms to predict is not enough as they do not make a difference in patient management, alter the disease outcome, or reduce necessary medical costs. He recommended
A science writer with data background and an interest in the current affair, culture,