Ensemble-learning-assisted exhaled gas disease analysis based on in-situ construction of MOF-derived MOx/GaN heterojunction sensor arrays

Human exhaled gas is rich in biomarker information that could be used for early diagnosis of disease. With the development of nanotechnology and the Internet of Medical Things (IoMT), AI-assisted nano gas sensor arrays as a non-invasive exhaled gas detection platform brings fascinating technological solutions to the field of breath detection. Herein, we designed a new heterojunction sensing array by anchoring n-GaN nanoparticles on MOF-derived p-MO_x porous nanosheets. The gas sensor arrays demonstrated remarkable response speed (10 s), excellent repeatability, and extreme anti-humidity with a lower detection limit of 1 ppb at room temperature. Energy band structure combined with density functional theory (DFT) calculations were used to analyze the entire gas sensing process. Furthermore, we developed a new breath detection device and successfully performed clinical patient exhaled gas detection. With the assistance of ensemble learning, the recognition accuracy of lung cancer patients and healthy volunteers can reach 95.8%. This work provides an innovative technology for the construction of heterojunction sensor arrays and exhaled gas detection device, which has a promising application prospect in the field of early disease diagnosis and IoMT.