Artificial intelligence-based chest X-ray image reading assistance has demonstrated its effectiveness in real-world clinical settings. For example, the detection rate of lung nodules used in lung cancer diagnosis based on AI programs showed an improved detection rate.

The research team led by Professors Ku Min-mo and Nam Ju-kang of the Radiology Department at Seoul National University Hospital (SNUH) produced such results by conducting a randomized clinical trial to know whether AI-based computer-aided diagnostic system (CAD) helps improve lung nodule detection in chest X-ray images.

The researchers carried out the trial to know whether AI-based CAD could improve lung nodular detection rate based on X-ray images of 10,476 people who conducted health checkups at SNUH from June 2020 to December 2021.

As a result, the detection rate of lung nodules in the AI-using group was 0.59 percent, 2.4 times higher than that of the non-using group’s 0.25 percent. The detection accuracy was also similar. In the CT test of positively tested patients, the pseudo request rate that found no nodules was 45.9 percent for the AI-using group and 56.0 percent for the non-using group.

The research team explained that using AI can improve the detection rate of nodules while suppressing unnecessary additional tests. However, these results were limited to cases where a radiologist did a final reading while referring to AI analysis. If AI alone reads them, the possibility of positive judgment increases even though there were no nodule increases, leading to unnecessary additional tests.

“We confirmed significant effects by applying AI models to medical images,” Professor Ku said. “It was meaningful that we confirmed medical professionals could find key opinions using AI.”        

The study result was published in the latest issue of Radiology.