At AIMed North America 2018, Dr. Christina Chen, nephrologist of Beth Israel Deaconess medical center, Harvard medical school expressed the privilege to work alongside with data scientists. Indeed, it has never been as challenging as to assemble experts with different agenda to work on the same project revolving around artificial intelligence (AI) in medicine.
Just data, the backbone of AI models, alone, both industrial and medical experts are already sharing a vast different opinion. The industry considered data as an asset with commercial values. In medicine, data is regarded as a commodity. Clinicians and researchers can’t wait to liberate data and have been urging for more open resources.
Fragmented views and separate developments may lead to the creation of AI models which clinicians do not have effective access to. Saurabh Gombar, clinical instructor of Stanford health care said by keeping experts in the loop and keep the developed AI models between them, it will increase the usability in the healthcare setting, making it easier to be introduced to the real World eventually.
Dr. Chen added, the same logic applies to AI model itself too. Until we have a finite set of data which can capture everything and human who will not instil bias into developing AI models, there is always a need to bring a human into the conversation because AI model cannot have solitude existence.
Gradual change in present policies
Collaborations do not limit between industrial partners and clinicians. Dale Van Demark, partner of McDemrmott Will & Emery LLP said medicine is probably the most regulated sector. We do not regulate nuclear power as much as we do for medicine, in every level and aspect. US is equipped with a system for dedicated individuals to deliver medical solutions all the time but regulations turn them towards marketing and reimbursement. These people have to pitch their products and services in a certain way, to obtain the Food and Drug Administration’s (FDA) approval.
Thus, instead of focusing on more pressing needs such as improving patient provider relationship or overcoming medical malpractice, companies are putting their resources on FDA approval. Ultimately, the FDA stamp means everything; no doctor will want to use it and patients will not be comfortable without it.
Van Demark stressed that the solution is not to do away with rules and regulations but to minimise the rigidity and show some flexibility towards new technology. Allison Kumar, chief executive officer and principle consultant of Arina Consulting, LLC added FDA is working to do away with the traditional medical devices based model and to include the machine learning based diagnostic device model. This, hopefully, will act as a breakthrough and shorten the normal regulatory approval time.
Begin with a more holistic education
Other speakers believe if AI and related new technology are integrated into the curriculum, future medicine will be less fragmented. Dr. Sharief Taraman cited there were 5310 journal articles published in 2017 which touched on machine learning (ML) and only 49 talked about ML and medical education. This suggests a disconnection between ML and present medical education.
“We need to start pushing policymakers to create databases for clinicians and empower patients to be part of it too. Education is not only limit to the medical community but everyone” said Dr. Taraman. His insight was supported by other panel speakers, including Jack Hidary, chairman of the Hidary foundation, who advocate sending doctors to code camp and to know how AI actually work from the insiders’ point of view.
“95 years ago, one need to know all kinds of mechanisms before they know what is going on with technology. Now, we have Github and all kinds of tools to learn how to code. There is never been a better time to learn about technology… it’s time to rethink the medical curriculum” Hidary said.
A science writer with data background and an interest in current affair, culture and arts; a no-med from an (almost) all-med family. Follow on Twitter.