Smartphone-based AI detects tuberculosis better than human specialists

As artificial intelligence (AI) technology keeps evolving, its capabilities are becoming increasingly impressive, including in the field of tuberculosis detection, where a new AI-powered lung ultrasound outperforms human experts.

Notably, the ULTR-AI suite analyzes images from portable, smartphone-connected ultrasound devices, with results exceeding the World Health Organization (WHO) benchmarks for pulmonary tuberculosis diagnosis, according to a report by News Medical published on April 14.

Indeed, this technology, demonstrated at ESCMID Global 2025, outperforms human experts by 9% in this area, presenting a sputum-free, quick, accessible, and scalable alternative for detecting tuberculosis, particularly important in the light of rising TB rates by 4.6% between 2020 and 2023.

Deploying the new AI capacities would help offset the high cost of chest X-ray equipment and a lack of trained radiologists that leads to substantial patient dropout in many high-burden countries. According to Dr. Véronique Suttels, the study’s lead author, this allows for test standardization and a reduction in operator dependency, helping patients reach a diagnosis faster and more efficiently, because:

“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.”

How smartphone-based AI TB detection system works

Specifically, the ULTR-AI system consists of three deep-learning models – ULTR-AI predicting TB directly from lung ultrasound images, ULTR-AI (signs) detecting ultrasound patterns as interpreted by human specialists, and ULTR-AI (max) using the highest risk score from both models to maximize precision.

As it happens, the researchers conducted the study at a tertiary urban center in Benin, West Africa, with ULTR-AI (max) demonstrating 93% sensitivity and 81% specificity, thus surpassing WHO’s target thresholds of 90% sensitivity and 70% specificity for non-sputum-based TB triage tests. In the words of Dr. Suttels:

“A key advantage of our AI models is the immediate turnaround time once they are integrated into an app. (…) This allows lung ultrasound to function as a true point-of-care test with good diagnostic performance at triage, providing instant results while the patient is still with the healthcare worker. Faster diagnosis could also improve linkage to care, reducing the risk of patients being lost to follow-up.”

Elsewhere, an international team of scientists has developed a novel AI model that can analyze microscopic images of cells and tissue to identify endometrial cancer with an impressive 99.26% accuracy rate, outperforming human-led methods that have an accuracy of around 78.91% to 80.93%.