Model outperforms radiologists for both recall and F1 score<\/p>\n<\/h3>\n
By Lori Solomon HealthDay Reporter<\/b>
\n<\/b><\/p>\n
MONDAY, Nov. 24, 2025 (HealthDay News) — An artificial intelligence (AI)-based model may aid assessment of suspected radiolucent foreign body aspiration on chest computed tomography, according to a study published online Nov. 10 in npj Digital Medicine<\/em>.<\/p>\n Xiaofan Liu, from the Huazhong University of Science and Technology in Wuhan, China, and colleagues developed a deep learning model integrating MedpSeg, a high-precision airway segmentation method, with a convolutional classifier to detect radiolucent foreign body aspiration, with training and validation occurring on three independent cohorts (more than 400 participants).<\/p>\n The authors report that the model showed consistent performance, with accuracies above 90 percent and balanced recall-precision metrics. The model outperformed expert radiologists in both recall (71.4 versus 35.7 percent) and F1 score (74.1 versus 52.6 percent) in a blinded independent evaluation cohort, suggesting the model’s potential to reduce missed cases (false negatives) and support clinical decision-making.<\/p>\n “The results demonstrate the real-world potential of AI in medicine, particularly for conditions that are difficult to diagnose through standard imaging,” lead author Yihua Wang, M.B.B.S., from the University of Southampton in the United Kingdom, said in a statement.<\/p>\n