Converge of coordinate attention boosted YOLOv5 model and quantum dot labeled fluorescent biosensing for rapid detection of the poultry disease.

• A new immunoassay using AI model was developed to rapidly detect candidiasis. • The YOLOv5 model was boosted by coordinate attention mechanism to detect pathogens quantitatively. • Transfer learning and multistage training strategy were used to train the model. • This proposed approach enabled the detection of Candida albicans as low as 3.5×102 CFU/mL in 45 min with an accuracy of 92.7%. Rapid detection of pathogens is of great significance to prevent the outbreak of poultry diseases. In this study, a new immunoassay based on coordinate attention boosted YOLOv5 model and quantum dot labeled fluorescence biosensing was developed to rapidly detect a common fungal infectious poultry disease, candidiasis. The target pathogen, Candida albicans , were specifically labeled by the immune quantum dots to form the fluorescent pathogen. Their fluorescent images were collected using a fluorescence microscope. The YOLOv5 model was boosted by coordinate attention mechanism for identifying the fluorescent pathogen in these images to obtain the quantitative results. With the help of transfer learning and multistage training, the intelligent model was successfully trained and applied. This proposed approach enabled the detection of Candida albicans as low as 3.5 × 102 CFU/mL in 45 min with an accuracy of 92.7 %, which was significantly higher than 87.2 % of YOLOv5 and 81.6 % of Faster RCNN. The new approach was featured with shorter detection time, lower cost and simpler operation, and showed its potential to provide a more efficient solution for pathogen screening. [ABSTRACT FROM AUTHOR]