U.K. Prime Minister Theresa May recently unveiled a “new weapon” in cancer research: artificial intelligence (AI).
She and medical experts say AI could prevent 22,000 cancer deaths a year by 2033. Pooled together and cross-referenced with national data, patients’ medical records, lifestyle habits, and genetic information would be used to help the computer algorithms spot cancer earlier.
In a speech about using science to transform health, May said:
“Late diagnosis of otherwise treatable illnesses is one of the biggest causes of avoidable deaths. And the development of smart technologies to analyze great quantities of data quickly and with a higher degree of accuracy than is possible by human beings opens up a whole new field of medical research.”
May refers to the 50,000 Britons diagnosed with lung, bowel, prostate, and ovarian cancer each year. If doctors use AI technologies to reduce late diagnosis by half, estimates U.K. charity Cancer Research, then 22,000 fewer people would die within 5 years of diagnosis. This “pioneering” tech could transform medicine around the world.
But does this apply to rarer cancers?
How AI-Driven Diagnosis Works
It takes time and specialized training to analyze medical images and data, especially now that clinical data grows at a rate beyond human ability to manage.
Only a year or so ago, the concept of using AI to manage this data was unheard of. Now 50 percent of healthcare organizations hope to deploy it in the next 5 years.
Powered by deep neural networks (complex mathematical systems that can learn on their own from vast amounts of data), AI can spot things humans don’t, make unbiased judgments, and ultimately analyze more data quicker. And the more data these trained computers take in, the smarter they become.
The healthcare industry uses AI for 2 key applications: to accelerate drug discovery and to improve cancer diagnosis. Over in the U.S., research into the latter field has exploded. The University of Southern California and MIT are using AI to diagnose breast cancer, Weill Cornell has discovered how to distinguish types of cancer from thousands of images of cells, and Stanford is distinguishing malignant skin lesions from benign ones.
The aim is to save lives and reduce healthcare costs. Cancer screenings not only take time but are also subject to error. Using AI to make more informed, accurate clinical decisions could save the U.S. $150 Billion annually. However, both here and in the U.K., challenges abound.
The Challenges: Scale and Rare Disease
Critics doubt May’s plans to deploy AI on a large scale. The U.K. has the technological capability to cut rates of cancer, but no system in place for “creating the right infrastructure within the health service, separating hype and genuine innovation and ensuring the public’s highly personal data is used responsibly,” according to the BBC.
The U.S. faces similar challenges on an even larger scale. On top of that, we don’t have the means to diagnose rare diseases.
Take mesothelioma, a rare form of cancer caused exclusively by exposure to asbestos. About 3,200 Americans are diagnosed with mesothelioma, but that’s a relatively small number compared to the 224,390 diagnosed with lung cancer. Due to the low incidence rate and lack of public awareness about the disease and how it’s caused, we have a serious lack of mesothelioma research.
Without research, we have no data. Without data – and masses of it – AI can’t work effectively. So where do rare cancers come into the picture?
The Future Looks Brighter with Research
For a long time, mesothelioma patients have had no possible medical solutions besides surgery. Immunotherapy drugs have made advancements, but even those only extend patients’ short survival rate so much. As with other cancers, successful treatment of mesothelioma relies heavily on early diagnosis.
Unfortunately, mesothelioma diagnosis takes too long. The latency period between first exposure to asbestos and first signs of disease can take anywhere from 20 to 50 years, and even then, the symptoms are often mistaken for a cold or the flu. Few people realize they were exposed to asbestos, because many men and women were first exposed to asbestos when asbestos companies hid those risks from the public.
AI is still in its infancy, but experts believe the healthcare industry is only a few years away from ubiquitous use. That gives us little time to dedicate critical funds to mesothelioma research, so that the people dying prematurely from asbestos injuries can make the most of AI, too.
Could we save any of the nearly 40,000 lives lost to asbestos-related diseases every year? We don’t yet know. But it’s clear AI could fundamentally change how we approach mesothelioma, one unfairly injured patient at a time.