A study published on August 21 in the European Heart Journal has presented an innovative method for the widespread detection of heart disease. It suggests that people should provide their doctors with selfies that can be analyzed by deep learning machines configured to predict the presence and the particular stage of the disease. This is an apt example of the growing importance of artificial intelligence (AI) in medicine. Here is a list of just some of the applications that have been in the news recently.
Early detection of Alzheimer's
Deep learning can play a major role in the early detection of Alzheimer's. Scans that measure the level of glucose in the brain can be used to diagnose this neurodegenerative disease, but the interpretation of these images is far from easy when Alzheimer's is in its initial stages. At the same time, early detection is crucially important if treatment is to be effective before the disease has gone too far. Conducted by researchers at the University of San Francisco, a pioneering 2018 study developed an AI algorithm that succeeded in detecting Alzheimer's in scan images some six years before it would have otherwise been diagnosed.
New drug discovery
Several companies are launching machine learning programs in the race to discover new drugs. The goal is to make use of vast amounts of biological data in the design of new more effective treatments. Most of these companies have their own proprietary methods and tools, but all of these programs are based on the same three-stage process: leveraging of AI's capacity to analyze vast datasets, the use of stem cells to model diseases so as to understand how and why they develop, and finally confirming findings with a combination of computer and laboratory experiments. The sector is still in its infancy, but is already attracting a lot of attention, as evidenced by the recent 143 million dollars raised by the Californian company Insitro.
AI is already used in the diagnosis and follow-up phases for several types of cancer: breast cancer, melanoma, and lung cancer. In July 2019, a study published in the American scientific journal Archives showed that Google's deep learning AI, LYNA, could detect breast cancer metastases more effectively than pathologists. The study proposed using the algorithm to improve diagnosis and reduce the number of false negatives, an idea that has since been adopted by the European Desiree project. With regard to melanomas, the French National Institute of Health and Medical Research Inserm found that a database of at least 50,000 images was required to train an algorithm to identify signs of the pathology.
According to a study by a team of scientists from the University of California, which was recently published in the journal Nature Medicine, it is now possible to diagnose diabetes with a smartphone. The algorithm the team created uses a smartphone camera to conduct a new and wholly unprecedented non-invasive test for the condition, which it successfully identified in more than 80% of cases.
What about covid-19?
AI is already playing a major role in the fight against the global pandemic. On July 15, researchers published a method to predict the worsening of symptoms in coronavirus patients in the journal Nature. The method, which was created by a deep learning machine working on data provided by some 3,000 Chinese patients, has now resulted in the development of a digital tool to orient new cases admitted to hospital on the basis of the risk that they will develop severe symptoms.