AI-Powered Lung Ultrasound Outperforms Human Experts in Tuberculosis Diagnosis

6% from 2020 to 2023. Early screening and rapid diagnosis are essential elements of the World Health Organization’s (WHO) ‘End TB Strategy.’ However, many high-burden countries face significant patient dropouts at the diagnostic stage due to the high costs of chest X-ray equipment and a shortage of trained radiologists. These challenges highlight the critical need for more accessible diagnostic tools. Now, a groundbreaking study presented at ESCMID Global 2025 has revealed that an AI-powered lung ultrasound outperforms human experts by 9% in diagnosing pulmonary TB.

The ULTR-AI suite, developed at the University of Lausanne (Lausanne, Switzerland), processes images from portable, smartphone-connected ultrasound devices. This technology offers a sputum-free, rapid, and scalable alternative for TB detection. The results surpass the WHO’s benchmarks for diagnosing pulmonary tuberculosis, representing a major breakthrough for accessible and efficient TB triage. The ULTR-AI suite consists of three deep learning models: ULTR-AI predicts TB directly from lung ultrasound images; ULTR-AI (signs) identifies ultrasound patterns interpreted by human experts; and ULTR-AI (max) optimizes accuracy by using the highest risk score from both models.

Conducted at a tertiary urban center in Benin, West Africa, the study involved 504 patients after exclusions, with 192 (38%) confirmed cases of pulmonary TB. Among the participants, 15% were HIV-positive, and 13% had a history of TB. A standardized 14-point lung ultrasound sliding scan protocol was used, with human experts interpreting images based on typical lung ultrasound findings. A single sputum molecular test served as the reference standard. The results of ULTR-AI (max) demonstrated 93% sensitivity and 81% specificity (AUROC 0.93, 95% CI 0.92-0.95), exceeding WHO’s target thresholds of 90% sensitivity and 70% specificity for non-sputum-based TB triage tests.

“The ULTR-AI suite leverages deep learning algorithms to interpret lung ultrasound in real time, making the tool more accessible for TB triage, especially for minimally trained healthcare workers in rural areas. By reducing operator dependency and standardizing the test, this technology can help diagnose patients faster and more efficiently,” said lead study author, Dr. Véronique Suttels. “Our model clearly detects human-recognizable lung ultrasound findings—like large consolidations and interstitial changes—but an end-to-end deep learning approach captures even subtler features beyond the human eye. Our hope is that this will help identify early pathological signs such as small sub-centimeter pleural lesions common in TB.”