摘要
随着计算机视觉和医学影像处理技术的发展,肺结节图像的分析和处理变得越来越重要。然而,由于肺结节图像的样本和质量有限,如何增加图像样本并保持其真实性成为一个关键问题。为增加肺结节图像样本数量并保持其真实性,研究提出了一种基于生成对抗网络的肺结节图像扩增方法,并使用均方误差(MSE)、结构相似度(SSIM)和峰值信噪比(PSNR)3个参数作为指标,对3种不同的生成对抗网络结构的应用效果进行对比,发现StyleGAN模型最佳。为验证所提出方法的有效性,设计了一个基于YOLO-V5的优化模型结构。实验结果表明,经过扩增的图像在肺结节检测任务中的精度从78.2%提高到了86.3%。
With the advancement of computer vision and medical image processing,the analysis and processing of pulmonary nodule images have become increasingly important.However,due to the limited samples and quality of pulmonary nodule images,increasing the sample size while maintaining their authenticity poses a critical challenge.In order to increase the lung nodule image samples and maintain their authenticity,we proposed a lung nodule image augmentation method based on Generative Adversarial Network(GANs).We used three metrics of Mean Square Error(MSE),Structural similarity(SSIM)and Peak Signal-to-Noise Ratio(PSNR)as indicators to compare the augmentation effects of three different generative adversarial network structures.The StyleGAN model was found to be the best model.To validate the effectiveness of the proposed method,an optimization model structure based on YOLO-V5 is designed.The experimental results showed that the precision of pulmonary nodule detection tasks improved from 78.2%to 86.3%after augmentation.
作者
张凌晓
杜轶琛
唐存东
ZHANG Ling-xiao;DU Yi-chen;TANG Cun-dong(School of Computer and Software,Nanyang Institute of Technology,Nanyang 473004,China;Automation Engineering College,Henan Polytechnic Institute,Nanyang 473000,China)
出处
《南阳理工学院学报》
2024年第6期44-49,共6页
Journal of Nanyang Institute of Technology
基金
河南省科技攻关项目(232102321069)
河南省高等学校重点项目(24B520027)
河南省科协科普项目(HNKP2024214)。
关键词
图像扩增
肺结节检测
医学图像处理
生成对抗网络
目标检测
image augmentation
pulmonary nodule detection
medical image processing
generative adversarial networks
object detection
