Detection of citrus Huanglongbing at different stages of infection using a homemade electronic nose system

Citrus has a long history of cultivation and is rich in nutrients, making it one of the most popular fruits in the world. Citrus Huanglongbing (HLB) is known as citrus "cancer", spread through citrus psyllids and grafting, spreading fast and destructive. No effective treatment has been found, and the spread of HLB can only be controlled by detecting and removing diseased trees. HLB has a long incubation period and is contagious during the incubation period. The leaf symptom presentation of HLB is complex, with varying symptoms at different stages of infection, making detection extremely difficult. Therefore, there is an urgent need for a detection method that does not depend on leaf symptoms to identify HLB at different stages of infection accurately. In this paper, we investigated the sensor response mechanism of HLB-specific volatile organic compounds (VOCs) due to the stress response of the plant and created an electronic nose detection system that is specific to HLB. Effective identification of healthy and symptomatic HLB samples was achieved by a commercial electronic nose with wide applicability. The indicator weighting analysis of the sensor's classification contribution was also performed. It was aiming at the lack of commercial electronic nose's ability to detect HLB samples at different infection stages, and considering the lack of specialized electronic nose equipment for HLB detection. Based on the analysis of indicator weights of commercial e-nose sensors and the change rule of specific VOCs, the selection of dedicated enose sensors was realized. The sensor reaction chamber was fabricated through structural design, 3D modeling, and fluid simulation design. In addition, a stable sensor reaction environment was constructed using a sophisticated gas circuit control scheme. Sample signals obtained from homemade electronic nose were analyzed using the best combination of feature sets and pattern recognition algorithms. The results showed an identification rate of 86.54% for samples targeting HLB in the undemonstrative phase, 93.51% for HLB samples with typical symptoms, and 93.44% for samples with mixed zinc deficiency and HLB symptoms. As a specialized electronic nose system for HLB detection, the recognition rate for HLB-positive samples for the three infection stages was 93.59%. The results of the study demonstrated the high sensitivity of the homemade gas-sensitive sensing device to the specific odors produced by HLB infection samples and the reliability of the identification of HLB samples at different symptomatic stages.