Artificial Intelligence (AI) has made many advances in medicine, health care, and pharma and biotech efficiency. In particular, AI has improved the ability of physicians to diagnose disease. In many cases, experts' demand exceeds the available supply, causing strain and delays in the medical field. According to the Institute of Medicine at the National Academies of Science, Engineering and Medicine, diagnostic errors contribute to approximately 10 percent of patient deaths. AI aims to mimic human cognitive functions and increase the availability of healthcare data. Not only has AI shifted diagnostics, but it aids in increasing drug development, speeding up clinical trials, personalizing treatments, and improving gene editing.
Image courtesy of The Verge.
How is AI being used to diagnose diseases?
Most of AI’s applications are seen through a few types: chatbots, oncology, and rare diseases. Chatbots are able to communicate with users and compare symptoms from a database of diseases. An example of this is Babylon, a UK based startup. In oncology, AI algorithms can detect skin cancer from images of skin lesions. Stanford University has generated a program of this sort, using deep convolutional neural networks. This is a Deep Learning algorithm that can assign weights and biases to various aspects of an image, thus being able to differentiate them from one another. The last type, rare diseases, involves correlating facial analysis and detecting phenotypes related to rare hereditary diseases. An example of this is the Face2Gene app, which is only available to trained clinicians to disprove false positives.
Image courtesy of Data Revenue.
The application of AI in diagnosing diseases is just in its beginning stages. More advanced systems, such as combining multiple data sources like genomics, patient data, MRI, and CT, are still in development. This will enable AI to recognize various diseases. Currently, many AI-driven analysis tools are restricted to a range of pathologies. This can lead to mistrust with physicians since AI may overlook or misdiagnose signs of diseases health practitioners have less experience with.
Physician vs Machine?
AI has a performance that is comparable to medical experts, especially in image recognition related fields. In a study published by The Lancet Digital Health, they found that in the past few years, AI has become extremely accurate. The researchers looked for studies that compared the diagnostic effectiveness of physicians and AI. The 14 studies they selected determined that AI can correctly diagnose disease in 87% of the cases compared to healthcare professionals who have an 86% accuracy rate. These statistics show that AI is merely comparable to human output and is unlikely to replace them in the workforce. What's more probable is that it will be used as an aiding tool to continue physician diagnoses' accuracy and time efficiency.
Within that handful of high quality studies, we found that deep learning could indeed detect diseases ranging from cancers to eye diseases as accurately as health professionals. But it’s important to note that AI did not substantially outperform human diagnosis.
- Professor Alastair Denniston, University Hospitals Birmingham
COVID-19 Detection and AI
Computed Tomography (CT) scans have become a useful tool in detecting COVID-19 associated lung disease. In a study conducted with the University of Central Florida, it was found that AI could help diagnose COVID-19 and overcome challenges of the testing process. The algorithm was found to have 90% accuracy, with the ability to discern positive cases 84% of the time and negative cases 93% of the time.
CT scans can even detect COVID-19 in people without symptoms; however, the technique isn't commonly used since scans look similar to influenza-associated pneumonia results. The algorithm developed by the University of Central Florida can overcome this by identifying COVID-19 cases while distinguishing them from influenza. The algorithm will be a handy tool for physicians to have while navigating the treatment of COVID-19.
COVID-19 observed from CT scans (European Journal of Radiology).
References
“AI Can Detect COVID-19 in the Lungs like a Virtual Physician, New Study Shows.” ScienceDaily, ScienceDaily, 30 Sept. 2020, https://www.sciencedaily.com/releases/2020/09/200930144426.htm
“AI Just as Good at Diagnosing Illness as Humans.” Medical News Today, MediLexicon International, https://www.medicalnewstoday.com/articles/326460#Biases-may-exaggerate-AI-performance
“Artificial Intelligence in Medicine: The Top 4 Applications.” RSS, https://www.datarevenue.com/en-blog/artificial-intelligence-in-medicine
Faggella, Daniel. “Machine Learning for Medical Diagnostics - 4 Current Applications.” Emerj, Emerj, 14 Mar. 2020, https://emerj.com/ai-sector-overviews/machine-learning-medical-diagnostics-4-current-applications/
Article Author: Jennifer Law
Article Editors: Victoria Huang, Edie Whittington