Bacterial Inhibition Zone Recognition Method Based on YOLO

Antimicrobial susceptibility testing involves placing antibiotic-containing paper disks on bacterial culture media surfaces. These drugs affect bacterial growth, resulting in inhibition zones that reflect the bacteria's sensitivity to the drug. Therefore, using digital means for intelligent identification of inhibition zone size is of great practical significance. This article proposes a deep learning model based on the YOLO target detection algorithm, capable of automatically identifying inhibition zones. Firstly, we load the pre-trained YOLO model and classify the output candidate boxes into several categories: petri dish, inhibition zone, and paper disk. This adaptation makes the YOLO model suitable for our task of inhibition zone recognition. Next, we use processed inhibition zone images to train the pre-trained YOLO model. We then evaluate the performance of multiple YOLO models and determine the optimal values for relevant hyperparameters to enhance overall performance. Experimental results show that, compared to other YOLO algorithms, the model trained using the YOLov9 algorithm achieves optimal performance in recognizing inhibition zones. Based on the recognition results, OpenCV tools are used to precisely calculate the diameter of inhibition zones and further assess antimicrobial efficacy. Finally, we developed an intelligent mobile application that allows researchers to automatically measure the diameter of inhibition zones and evaluate antibacterial capabilities using mobile devices. This application enables more efficient antimicrobial susceptibility analysis and accelerates antibacterial drug research.