Humans excel in interpreting chest X-rays compared to AI.

Humans excel in interpreting chest X-rays compared to AI.

AI Tools Can’t Fully Replace Radiologists in Diagnosing Lung Diseases, Study Says


AI tools may boost radiologists’ confidence in their diagnoses, but they can’t be solely relied upon to identify common lung diseases on chest X-rays, according to a new study. This research, published in Radiology, compared the performance of 72 radiologists against four commercially available AI tools using over 2,000 chest X-rays.

“Chest radiography is a common diagnostic tool, but significant training and experience is required to interpret exams correctly,” said lead researcher Dr. Louis Plesner, resident radiologist and PhD fellow in radiology at Herlev and Gentofte Hospital in Copenhagen, Denmark. While AI tools may assist radiologists in interpreting chest X-rays, their real-life diagnostic accuracy remains unclear.

The study focused on three common findings in the X-rays: airspace disease, pneumothorax, and pleural effusion. The AI tools demonstrated sensitivity rates ranging from 72% to 91% for airspace disease, 63% to 90% for pneumothorax, and 62% to 95% for pleural effusion. However, radiologists outperformed AI in accurately identifying the presence and absence of these lung diseases.

“The AI tools showed moderate to high sensitivity comparable to radiologists for detecting airspace disease, pneumothorax, and pleural effusion on chest X-rays,” said Dr. Plesner. “However, they produced more false-positive results than the radiologists, particularly when multiple findings were present and for smaller targets.”

For pneumothorax, the probability that patients with a positive screening result truly had the disease ranged from 56% to 86% for the AI systems, compared to 96% for the radiologists. Dr. Plesner highlighted AI’s shortcomings in identifying airspace disease, with positive predictive values ranging between 40% and 50%. In certain cases, the AI predicted the presence of the disease when none was actually present.

While AI systems excel at detecting disease, they struggle to identify the absence of disease, especially in complex chest X-rays. “Too many false-positive diagnoses would result in unnecessary imaging, radiation exposure, and increased costs,” Dr. Plesner cautioned.

It’s important to note that prior studies claiming AI superiority over radiologists often lacked access to a patient’s clinical history and previous imaging studies. In real-world practice, a radiologist’s interpretation of an imaging exam involves synthesizing these variables along with the image itself.

Although AI tools have their advantages, they can’t replace the expertise and judgment of radiologists when it comes to accurately diagnosing lung diseases. Therefore, while AI may assist radiologists, these tools should not be solely relied upon in clinical practice. Further testing and evaluation are needed to determine the optimal role of AI in radiology departments.

More information: The American Hospital Association has additional information on using AI in diagnosis and care.

Sources: – Radiology, news release, Sept. 26, 2023 – Image Source

SLIDESHOW: Heart Disease: Causes of a Heart Attack