Innovative Point-of-Care Breath Test for Early Asthma Diagnosis: Integration of Deep Neural Network Technology and Nanosensors

<bold>Background:</bold> Asthma diagnosis requires meticulous clinical history assessment, prompting the need for a non-invasive, user-friendly diagnostic tool. Hence, we pioneered a point-of-care breath test for early asthma diagnosis using deep neural network (DNN) technology and nanosensors. <bold>Methods:</bold> Voluntary organic compounds (VOCs) biomarkers, capnographic waveforms, and clinical lung function parameters were collected in a blinded manner from a cohort of 20 participants, evenly divided between 10 individuals diagnosed with asthma and 10 without respiratory conditions. End-tidal breath samples were analysed with nanosensors, measuring 13 distinct parameters. A calibrated VOC sensor and standardized breath collection were used. The levels of VOC biomarkers were elevated in asthma patients compared to healthy volunteers, as illustrated in Fig. 1A. These findings highlight the stability and reproducibility of the results. Subsequently, a 13-layer multiclass classification DNN model, trained with 758 data points comprising biomarkers and lung function parameters, adopted a big data approach. <bold>Results:</bold> Our model achieved a 93% accuracy rate in predicting and diagnosing asthma, with an AUC of 0.95 (Fig. 1B). <bold>Conclusion:</bold> Exhaled biomarkers hold promise in differentiating asthma patients from healthy individuals, aiming to improve asthma diagnosis.