基于改进YOLOv5的荧光图像细胞智能检测研究

为解决人工对荧光原位杂交(Fluorescence In Situ Hybridization, FISH)荧光图像进行结果判读存在的效率低、劳动强度大等问题，针对FISH荧光图像细胞智能检测提出一种融合空域图像增强的改进YOLOv5算法。算法在原始YOLOv5神经网络模型基础上，加入了空域图像增强模块，并选择了模块最佳增强系数，扩大了模型对荧光图像的对比度适应范围，提高了模型的特征提取能力和细胞检测准确率。实验结果显示，改进YOLOv5模型的平均精度均值(Mean Average Precision, mAP)为0.983,达到了比原始模型更优的训练效果和收敛速度，并且，改进YOLOv5模型的细胞识别率达到91.65%,比原始YOLOv5模型提升了9.19%。将细胞智能检测算法嵌入自主开发的荧光图像智能检测软件，结合荧光点检测算法，可给出有效判读结果。

Intelligent Detection of Cells in Fluorescence Images Based on Improved YOLOv5

To solve the problems of low efficiency and high labor intensity in manual interpretation of FISH (Fluorescence In Situ Hybridization) fluorescence images, an improved YOLOv5 algorithm that integrates spatial image enhancement is proposed for intelligent cell detection in FISH fluorescence images. On the basis of the original YOLOv5 neural network model, the algorithm added a spatial image enhancement module, and the optimal enhancement coefficient of this module was selected. This module expanded the contrast adaptation range of the model to fluorescence images, and improved the feature extraction ability and cell detection accuracy of the model. The experimental results show that the mAP (Mean Average Precision) of the improved YOLOv5 model is 0.983, which achieves better training performance and convergence speed than the original model. Furthermore, the improved YOLOv5 model achieves a cell recognition rate of 91.65%, which is 9.19% higher than that of the original YOLOv5 model. Embedding the intelligent cell detection algorithm into the self-developed fluorescence image intelligent detection software, combined with fluorescence point detection algorithm, it can give effective interpretation results.