结合潜在扩散模型和U型网络的HIFU治疗目标区域提取  

Combination of latent diffusion and U-shaped networks for HIFU treatment target region extraction

作　　者：翟锦涛 王润民 李昂 田峰 龚瑾儒 钱盛友 邹孝 Zhai Jintao;Wang Runmin;Li Ang;Tian Feng;Gong Jinru;Qian Shengyou;Zou Xiao(School of Physics and Electronics,Hunan Normal University,Changsha 410081,China;Key Laboratory of Physics and Devices in Post-Moore Era,College of Hunan Province,Changsha 410081,China;School of Information Science and Engineering,Hunan Normal University,Changsha 410081,China)

机构地区：[1]湖南师范大学物理与电子科学学院,长沙410081 [2]后摩尔时代物理与器件湖南省普通高等学校重点实验室,长沙410081 [3]湖南师范大学信息科学与工程学院,长沙410081

出　　处：《中国图象图形学报》2024年第5期1291-1306,共16页Journal of Image and Graphics

基　　金：国家自然科学基金项目(12274200,61502164);湖南省教育厅科研基金项目(21A0052,22B0036);湖南省自然科学基金项目(2020JJ4057);湖南省社会科学基金项目(2021WLH35);长沙市自然科学基金项目(kq2202239)。

摘　　要：目的 由于数据采集限制和隐私保护造成高强度聚焦超声(high intensity focused ultrasound, HIFU)治疗超声监控图像数据量过少,导致现有的强监督分割方法提取治疗目标区域不佳。因此,提出了一种结合潜在扩散模型(latent diffusion)和U型网络的HIFU治疗目标区域提取方法。方法 生成阶段利用潜在扩散模型和自动筛选模块,实现超声监控图像数据的扩充。目标区域提取阶段提出新型U型分割网络(novel U-shaped segmentation network,NUNet),在编码器端结合空洞空间金字塔池化(atrous spatial pyramid pooling, ASPP),扩大网络的感受野;设计双注意力跳跃连接模块(dual attention skip connection,DAttention-SK),降低边缘纹理信息丢失的风险;引入多交叉熵损失提高网络的分割性能。结果 实验结果表明,与其他生成模型相比,本文使用潜在扩散模型生成的超声监控图像在FID(Fréchet inception distance)和LPIPS(learned perceptual image patch similarity)上获得更优的指标(分别为0.172和0.072);相较于先进的PDF-UNet(U-shaped pyramid-dilated network),在HIFU临床治疗子宫肌瘤超声监控数据集中,本文分割算法的MIoU(mean intersection over union)和DSC(Dice similarity coefficient)分别提高了2.67%和1.39%。为进一步探讨所提算法的泛化性,在乳腺超声公共数据集(breast ultrasound images dataset,BUSI)上进行了验证。相较于M2SNet(multi-scale in multi-scale subtraction network),本文算法MIoU和DSC分别提升了2.11%和1.36%。结论 本文算法在一定程度上解决了超声监控图像中数据量过少的问题,实现对监控超声图像中目标区域的精确提取。代码开源地址为https://github.com/425877/based-on-latent-diffusion-model-for-HIFU-treatment-target-region-extraction。Objective In high intensity focused ultrasound(HIFU)treatment,the target area contains a large amount of pathological information;thus, the target area must be accurately located and extracted by ultrasound monitoring images.As biological tissues and target regions change their relative positions during treatment, the location of the treatment areamay also change. At the same time, the diversity of diseases, the variability of tissues, and the complexity of target shapespose certain challenges for target region extraction in ultrasound medical images. Nevertheless, computers can useadvanced image processing and analysis algorithms, combined with big data and machine learning methods, to identify andlocate target areas quickly and accurately, providing a reliable basis for quantitative clinical analysis. Traditional imagesegmentation algorithms mainly include methods, such as threshold segmentation, edge detection, and region growing.However, these methods still have some limitations and are sensitive to the complexity of ultrasound images, noise, andother image quality issues, resulting in poor accuracy and robustness of segmentation results. Meanwhile, traditional meth⁃ods usually require manual selection of parameters, which limit the adaptive and generalization capabilities of the methods,and have a strong dependence on different images. In recent years, deep learning-based methods have attracted widespreadattention and made remarkable progress in the field of medical image segmentation. Most of the methods are performedunder strong supervision, yet this type of training requires a large amount of data as support for improved prediction. Theamount of data in HIFU therapy ultrasound surveillance images is too small due to patient privacy, differences in acquisi⁃tion devices, and the need for manual labeling of target areas by specialized physicians. It causes the network not to beadequately trained, making the segmentation results poor in accuracy and robustness. Therefore, this study proposed amethod for e

关 键 词：高强度聚焦超声(HIFU) 图像分割 图像生成 损失函数 潜在扩散模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]