Zhilin Han, Yuyang Zhang, Wenlong Ding, Xiaoting Zhao, Bingzhen Jia, Tingting Liu, Liang Wan, Zhiheng Xing
Sci Data . 2025 Mar 29;12(1):533. doi: 10.1038/s41597-025-04838-8.
The increasing global incidence of nontuberculous mycobacterial (NTM) pulmonary disease highlights the need for rapid diagnostic methods to guide timely treatment and prevent antibiotic misuse. While bacterial culture remains the gold standard for diagnosis, its extended turnaround time compromises clinical decision-making. Computed tomography (CT), with its high sensitivity for lung lesions and rapid imaging capabilities, has emerged as a critical diagnostic tool. AI-assisted CT interpretation shows particular promise for improving NTM detection, yet progress has been hindered by limited datasets due to disease rarity. We address this gap by introducing the first comprehensive CT dataset combining 430 NTM and 871 tuberculosis cases, supplemented with clinical parameters including demographics, symptoms, and mycobacterial species data. This resource aims to catalyze AI algorithm development for differential diagnosis, ultimately enhancing precision in NTM management through advanced machine learning applications.