“AI medical diagnosis accuracy is similar to that of humans”

Artificial intelligence is also expected to be used in the medical field. Studies are often reported that AI, which is learned based on vast images and data, is superior to human doctors. However, most of the papers that study the application of AI to the medical field are inadequate, and the results of investigations that the accuracy of diagnosis between AI and humans do not change in reality draw attention.

Medical diagnostic research by AI is in rapid progress in recent years, and there are many studies that have concluded that AI examines with higher precision than humans. For example, there is a report that the AI for lung cancer detection developed by Google boasts a higher detection rate than human doctors. In addition, because AI diagnoses with higher accuracy than humans, it is argued that AI diagnoses and does not require unnecessary resection surgery due to diagnosis errors.

The research team at University Hospitals Birmingham (UHB) investigated research papers related to deep learning published since 2012. Although there are more than 20,000 research papers on AI diagnosis, it is said that only 14 of them were found in comparative studies that provided high-quality data and showed images that seemed to analyze AI to human experts.

The research team compiled the most promising results of 14 studies to calculate the accuracy of AI to accurately diagnose human diseases. It is said that while a human doctor performed an accurate examination with 86% accuracy, the accuracy of AI was 87%. In addition, the accuracy of disease diagnosis was 91% for humans, compared to 93% for AI. In other words, there was no significant difference when comparing the accuracy of AI diagnosis with human doctors, considering only the studies with substantial content. However, the research team revealed that caution is needed as this result was also derived from a small number of studies.

One expert points out that AI application research is lacking in the medical field. This study proves that the quality level of most studies is unfortunate by the hype about AI in medicine. Say there is.

Of course, it is positive about the potential of AI in the medical field. AI can function as a diagnostic tool and help cope with medical imaging tests. If it is difficult to secure human resources to perform image diagnosis, it is expected that AI will also be available here. It is clear that AI will be important in the future even in the medical field, but it is pointed out that target testing of AI is needed. He says it’s also important to understand the reasons why AI makes the wrong assessment. Related information can be found here .