SHOCKING STUDY REVEALS: Radiologists Outperform AI in Detecting Lung Diseases!
It turns out that radiologists have the upper hand when it comes to identifying common lung diseases compared to artificial intelligence (AI) systems, according to a groundbreaking study published in the prestigious journal Radiology. The study, led by Dr. Louis L. Plesner from Herlev and Gentofte Hospital in Denmark, compared the performance of four commercially available AI tools with a pool of 72 radiologists in interpreting over 2,000 chest X-rays. And the results are jaw-dropping!
The study focused on three common lung diseases: airspace disease, pneumothorax (collapsed lung), and pleural effusion (water buildup around the lungs). While the AI tools showed moderate to high sensitivity in detecting these diseases, they produced more false-positive results than the radiologists. In fact, the AI systems predicted disease where none was present five to six out of 10 times! Imagine the consequences of unnecessary imaging, radiation exposure, and increased costs!
But that’s not all! The AI tools also struggled when multiple findings were present and when dealing with smaller targets. Their performance paled in comparison to that of the seasoned radiologists. And let’s talk about positive predictive values, which measure the probability of patients with a positive screening test truly having the disease. The AI systems failed miserably in this aspect, with values ranging from 56 to 86% for pneumothorax, while the radiologists achieved an impressive 96%!
Dr. Plesner stressed the importance of balancing the ability to find and exclude disease, avoiding both overlooked diseases and overdiagnosis. While AI systems may excel at finding disease, they fall short in identifying the absence of disease, especially in complex cases. The next generation of AI tools may improve if they can synthesize patient history and previous imaging studies, but no such systems exist yet.
So, what’s the bottom line? Radiologists outshine AI in real-life clinical scenarios, where the variety of patients poses challenges that AI cannot handle. While AI can provide a second opinion and boost radiologists’ confidence, it should not be autonomous in making diagnoses.
But what do you think? Can AI ever match the expertise of radiologists? Share your thoughts below!
IntelliPrompt curated this article: Read the full story at the original source by clicking here