Non-Ionizing 6G Signals Accurately Classify Lung Diseases Without Contact

Air pollution, smoking, infections, occupational exposures, psychosocial stress, and genetic factors have contributed to respiratory diseases becoming a significant global health burden. The World Health Organization reported that respiratory diseases were the third leading cause of death worldwide in 2019, accounting for approximately 3.8 million deaths annually. 

Conventional diagnostic methods, such as chest X-rays, computed tomography (CT) scans, spirometry, and sputum analysis, are effective but are often expensive, invasive, and require specialized equipment and trained personnel. Some techniques involve exposure to ionizing radiation or need close physical contact, which is less desirable in a post-COVID-19 healthcare environment. To overcome these challenges, the researchers have developed a new contactless screening technology that uses non-ionizing 6G/Wi-Fi radio waves in combination with artificial intelligence (AI).

The researchers investigated the possibility of diagnosing lung diseases by analyzing changes in breathing movements reflected in radio signals from the chest. Multi-carrier orthogonal frequency division multiplexing (OFDM) radio signals operating at a center frequency of 5.23 GHz were transmitted toward the chest of participants. The breathing patterns that each respiratory condition influences the amplitude, phase, and frequency of the reflected signals. The dataset was termed OFDM-Breathe and was collected in a hospital environment from 220 individuals, including 190 patients and 30 healthy controls. Of the patients, 45 had asthma, 43 had chronic obstructive pulmonary disease (COPD), 41 had tuberculosis (TB), 30 had interstitial lung disease (ILD), and 31 had pneumonia (PN). The dataset comprised 26,760 seconds of radio signal recordings across 64 frequencies.

The experiment system utilized two software-defined radios with 64OFDM subcarriers, quadrature phase-shift keying (QPSK) modulation, 1000 samples/sec sampling rate, and a gain of 40 dB on transmission and reception. The captured signals were treated to obtain channel frequency response information, which was subsequently denoised, segmented, and pre-processed for analysis. Several machine learning algorithms and deep learning systems, such as support vector machines, multi-layer perceptrons, transformers, long short-term memory networks, and convolutional neural networks (CNNs), were trained to classify respiratory conditions.

A standard convolutional neural network achieved the best performance and obtained 98% accuracy in distinguishing between the five lung diseases and healthy people among all of the tested models. The CNN also exhibited high precision, recall, and F1-scores across all disease classes. It is important to note that all deep learning models identified normal breathing patterns with 100% recall, suggesting strong reliability in detecting abnormal respiratory activity. Ablation studies showed that high diagnostic performance could be achieved using fewer frequencies: the system maintained 96% accuracy using only eight subcarriers (12.5% of the total bandwidth), allowing the remaining 87.5% of the bandwidth to support standard 6G/Wi-Fi data communication.

This study demonstrates that contactless, radio-sensing (non-ionizing), when used in combination with AI, can be trusted to screen for asthma, chronic obstructive pulmonary disease (COPD), interstitial lung disease (ILD), pneumonia, and tuberculosis. The proposed method does not need physical contact, radiation, or costly imaging tools, unlike traditional diagnostic tools. Its ability to operate with limited bandwidth while maintaining high accuracy makes it particularly suitable for future 6G integrated sensing and communication systems.

This strategy could enable real-time respiratory screening in hospitals, homes, and remote or resource-limited settings, supporting early detection and promoting health equity. Although not intended to replace established clinical diagnostics, this technology represents a valuable complementary tool that may reduce diagnostic delays and expand access to respiratory health monitoring worldwide.  

Reference: Buttar HM, Rahman MMU, Nawaz MW, et al. Non-contact lung disease classification via orthogonal frequency division multiplexing–based passive 6G integrated sensing and communication. Commun Med. 2026;6:9. doi:10.1038/s43856-025-01181-2