Man Vs Machine Artificial Intelligence Outperforms Doctors at Diagnosing Diseases
AI Beats Experts at Diagnosing Diseases

There is no doubt that artificial intelligence or AI has immense potential. In the next few years, it will revolutionize every field of our life, including healthcare. The technology has sent vast waves across the sector, powered by the growing availability of healthcare data and the fast progress of analytic procedures. Due to its power of transforming healthcare, there have been active discussions of whether AI will take the place of physicians/doctors in the future. One thing is for sure that even if AI will not replace physicians/doctors, it can surely help them tackle key issues in the healthcare system. It can help make better clinical decisions or replace human judgment. AI tools find use in major disease areas such as cancer, neurology, and cardiology.

Recently, researchers have shown that AI is better at detecting skin cancer than experienced dermatologists. A team from Germany, the United States, and France taught an artificial intelligence system called deep learning convolutional neural network (CNN) to differentiate between dangerous skin lesions and benign ones by showing it over 100,000 images of malignant melanomas, the deadliest form of skin cancer, as well as nevi, benign moles. Inspired by biological methods, the CNN is an artificial neural network capable of learning quickly from images that it sees and teaching itself from what it has learned to improve its performance.

The performance of the machine was compared with that of 58 international dermatologists. The researchers found that CNN missed fewer melanomas and misdiagnosed benign moles than the group of dermatologists. The dermatologists accurately identified about 86.6% of melanomas and about 71.3% of nevi. Nevertheless, the CNN correctly identified 95% of melanomas.

“The CNN missed fewer melanomas, meaning it had a higher sensitivity than the dermatologists, and it misdiagnosed fewer benign moles as malignant melanoma, which means it had a higher specificity; this would result in less unnecessary surgery,” said Professor Holger Haenssle from the University of Heidelberg. “When dermatologists received more clinical information and images at level II, their diagnostic performance improved. However, the CNN, which was still working solely from the dermoscopic images with no additional clinical information, continued to outperform the physicians’ diagnostic abilities,” he said.

AI has also proven to be useful in other areas of healthcare such as cardiology. A team of researchers from the John Radcliffe Hospital in Oxford, England created an AI diagnostics system known as Ultromics which can diagnose heart disease more accurately than doctors. The system was trained using the heart scans of 1,000 patients treated by the company’s chief medical officer, Paul Leeson, as well as information about if those patients suffered heart problems. The system has been tested in several clinical trials, and it was found that it had greatly outperformed human cardiologists.

According to a new study from Oregon Health and Science University, an AI system can diagnose a childhood blindness disease called retinopathy of prematurity (ROP) better than most experts. ROP is a disease that can cause blindness in premature babies. To diagnose it, physicians require to examine the back of the eye and check if the blood vessels are extremely dilated or wiggly. It’s difficult to diagnose. And if the disease is left undiagnosed or not detected at an early stage, a child could be blind for the rest of his/her life. However, scientists from OHSU and Massachusetts General Hospital developed a computer system which compares an image of a patients’ retina against several healthy and ROP images and then uses a system of deep-learning to deliver a diagnosis. The algorithm correctly diagnosed the condition 91% of the time compared to a team of eight experts who diagnosed the condition 82% of the time.