基于DYOLO的超声图像肾脏疾病自动检测

王子昊, 龚骁晨

中国医疗设备 ›› 2025, Vol. 40 ›› Issue (1) : 34-40.

PDF(2596 KB)
PDF(2596 KB)
中国医疗设备 ›› 2025, Vol. 40 ›› Issue (1) : 34-40. DOI: 10.3969/j.issn.1674-1633.20240748
研究论著

基于DYOLO的超声图像肾脏疾病自动检测

  • 王子昊, 龚骁晨
作者信息 +

Automatic Detection of Kidney Disease in Ultrasound Images Based on DYOLO

  • WANG Zihao, GONG Xiaochen
Author information +
文章历史 +

摘要

目的 以超声成像技术为基础,提出一种基于目标检测模型YOLO的肾脏疾病检测技术,以快速有效地对肾脏疾病进行检测,改善超声图像检测效果。方法 本研究将卷积与YOLO融合于同一个训练框架并构建融合DYOLO模型。考虑传统卷积特异性的限制问题,采用空间变换模型改进模型,同时引入CSPDarknet53模块提取图像特征,从而实现对肾脏形状大小的自动调整与识别。结果 在病理图像采样的效果对比中,卷积模型、YOLOv3模型、YOLOv4模型以及改进DYOLO模型图像的采样准确度分别为68.65%、82.65%、90.65%与95.65%。在复杂肾脏数据测试中,改进DYOLO模型在迭代600次后趋于收敛,最佳平均精度均值为95.23%,表现效果最佳。结论 改进DYOLO模型具有出色的病理特征检测能力,对肾脏疾病的检测具有重要参考意义。

Abstract

Objective Based on ultrasound imaging technology, to propose a kidney disease detection technique using the YOLO object detection model to quickly and effectively detect kidney diseases, improving the detection performance of ultrasound images. Methods The convolution was combined with YOLO in the same training framework and a fused DYOLO model was built in this research. Considering the limitations of traditional convolution specificity, a spatial transformation model was used to improve the model. At the same time, the CSPDarknet53 module was introduced to extract image features, thus achieving automatic adjustment and recognition of kidney shape and size. Results In the comparative effectiveness analysis of pathological images, the image sampling accuracy of the convolution model, YOLOv3 model, YOLOv4 model, and improved DYOLO model were 68.65%, 82.65%, 90.65%, and 95.65%, respectively. In complex kidney data testing, the improved DYOLO model converged after 600 iterations, with an optimal mean average precision of 95.23%, demonstrating the best performance. Conclusion The improved DYOLO model has excellent pathological feature detection capabilities, which is of significant reference value for the detection of kidney diseases.

关键词

DYOLO模型;超声图像;肾脏疾病;CSPDarknet53;卷积;图像特征;空间变换

Key words

DYOLO model; ultrasound images; kidney disease; CSPParknet53; convolution; image features; spatial transformation

引用本文

导出引用
王子昊, 龚骁晨. 基于DYOLO的超声图像肾脏疾病自动检测[J]. 中国医疗设备, 2025, 40(1): 34-40 https://doi.org/10.3969/j.issn.1674-1633.20240748
WANG Zihao, GONG Xiaochen. Automatic Detection of Kidney Disease in Ultrasound Images Based on DYOLO[J]. China Medical Devices, 2025, 40(1): 34-40 https://doi.org/10.3969/j.issn.1674-1633.20240748
中图分类号: R144    R445   

参考文献

[1] 魏雪, 章如松, 李颖, 等. 超声波处理快速石蜡制片在供肾零 点活检病理诊断中的应用与比较[J]. 临床与实验病理学杂志, 2022, 38(5): 629-631.
[2] 刘奇, 赵丽霞, 郑曙光, 等. 基于DYOLO神经网络的超声图像 肾脏检测[J]. 计算机工程, 2021, 47(7): 307-313.
Liu Q, Zhao LX, Zheng SG, et al. Kidney detection using ultrasound image based on DYOLO neural network[J]. Comput Eng, 2021, 47(7): 307-313.
[3] 伊佳琪, 孙晓刚. 基于卷积神经网络的轮对超声波探头倾斜检 测[J]. 计算机应用, 2021, 41(z2): 280-285.
Yi JQ, Sun XG. Wheelset ultrasonic probe tilt detection based on convolutional neural network[J]. J Comput Appl, 2021, 41(z2): 280-285.
[4] 包泽伟, 林建原, 张瑱, 等. 超声波辅助酶解山茶籽油提取工艺优 化及人工神经网络搭建[J]. 中国粮油学报, 2022, 37(9): 204-209.
Bao ZW, Lin JY, Zhang Z, et al. Optimization of ultrasonie assisted enzymatic hydrolysis of camellia seed oil extraction process and construction of artificial neural network[J]. J Chin Cereals Oils Assoc, 2022, 37(9): 204-209.
[5] 董淼, 牛蔓丽, 蒋纬昌. 基于Deeplabv3+的街景图像语义分割 技术研究[J]. 现代科学仪器, 2023, 40(2): 160-165.
Dong M, Niu ML, Jiang WC. Research on semantic segmentation of streetscape image based on deep labv3+[J]. Mod Sci Instrum, 2023, 40(2): 160-165.
[6] 熊杰, 魏勇, 严丹. 基于BP神经网络的超声波温湿度补偿算法 研究与应用[J]. 现代电子技术, 2020, 43(9): 113-116.
Xiong J, Wei Y, Yan D. Research and application of ultrasonic temperature-humidity compensation algorithm base on BP neural network[J]. Mod Electr Tech, 2020, 43(9): 113-116.
[7] Weerasinghe NH, Lovell NH, Welsh AW, et al. Multi-parametric fusion of 3D power Doppler ultrasound for fetal kidney segmentation using fully convolutional neural networks[J]. IEEE J Biomed Health Inform, 2020, 25(6): 2050-2057.
[8] 马立勇, 董梦超, 李广涵, 等. 基于剪切波弹性成像和卷积神经 网络建立深度学习模型预测肾脏病变[J]. 中国医学影像技术, 2021, 37(6): 919-922.
Ma LY, Dong MC, Li GH, et al. Predicting renal diseases with deep learning model based on shear wave elastography and convolutional neural network[J]. Chin J Med Imaging Technol, 2021, 37(6): 919-922.
[9] 谢杨, 马迎春, 杨吉刚, 等. 基于多层感知神经网络的肾小球 滤过率评估模型的开发和验证[J]. 中华肾脏病杂志, 2022, 38(5): 369-378.
Xie Y, Ma YC, Yang JG, et al. Development and validation of multilayer perceptual neural network in glomerular filtration rate evaluation[J]. Chin J Nephrol, 2022, 38(5): 369-378.
[10] 陈晗, 郑彩仙. 基于改进卷积神经网络图像处理的医疗设备 质量控制研究[J]. 现代科学仪器, 2023, 40(1): 166-170.
Chen H, Zheng CX. Research on power grid technical transformation project group management based on adaptive KFCM[J]. Mod Sci Instrum, 2023, 40(1): 166-170.
[11] Patil S, Choudhary S. Deep convolutional neural network for chronic kidney disease prediction using ultrasound imaging[J]. Bio-Algorithms Med-Syst, 2021, 17(2): 137-163.
[12] 钱程丽, 朱勤, 陈洪宇. 人工神经网络在肾脏疾病方面的应用 及展望[J]. 中国中西医结合肾病杂志, 2021, 22(3): 269-271.
[13] 孙晓晗, 孔祥勇, 吴滢, 等. 基于卷积神经网络的肾小球病理图 像分类算法[J]. 中国医学物理学杂志, 2022, 39(10): 1313-1320.
Sun XH, Kong XY, Wu Y, et al. Glomerular pathological image classification algorithm based on convolutional neural network[J]. Chin J Med Phys, 2022, 39(10): 1313-1320.
[14] 冯蕾, 黄菊秀, 赵冉冉. 深度卷积神经网络图像超分辨率重建 方法研究[J]. 现代科学仪器, 2022, 39(1): 205-208.
Feng L, Huang JX, Zhao RR. Image super-resolution restoration instance-based learning and iterative kernel regression[J]. Mod Sci Instrum, 2022, 39(1): 205-208.
[15] 马紫童, 王雪帆, 唐秀凤, 等. 基于神经网络模型评价淫羊 藿女贞子配伍对大鼠心肺衰老的影响[J]. 吉林中医药, 2021, 41(2): 238-243.
Ma ZT, Wang XF, Tang XF, et al. Evaluation of the effect of compatibility of epimedium herb and glossy privet fruit on aging of heart and lung in rats based on neural network model[J]. Jilin J Chin Med, 2021, 41(2): 238-243.
[16] 梁利渡, 张浩杰, 鲁倩, 等. aFaster RCNN: 一种基于平扫 CT的多疾病阶段胰腺病灶检测模型[J]. 南方医科大学学报, 2023, 43(5): 755-763.
Liang LD, Zhang HJ, Lu Q, et al. Advanced Faster RCNN: a non-contrast CT-based algorithm for detecting pancreatic lesions in multiple disease stages[J]. J South Med Univ, 2023, 43(5): 755-763.
[17] 李宗霖, 赵峥, 连世东. 基于SPECT全身骨扫描的YOLOv5x深度学习网络模型诊断良、恶性骨病灶[J]. 中国医学影像技 术, 2023, 39(12): 1867-1871.
Li ZL, Zhao Z, Lian SD. YOLOv5x deep learning network model based on SPECT whole body bone scanning for diagnosing benign and malignant bone lesions[J]. Chin J Med Imaging Technol, 2023, 39(12): 1867-1871.
[18] 俞永伟. 改进型YOLOv5s网络在胆囊超声图像检测中的应 用[J]. 中国医疗设备, 2023, 38(5): 99-104.
Yu YW. Application of improved YOLOv5s network in gallbladder ultrasound image detection[J]. China Med Devices, 2023, 38(5): 99-104.
[19] 王正业, 热娜古丽·艾合麦提尼亚孜, 王晓荣, 等. 肝囊型包 虫病超声图影像区域分割算法研究[J]. 中国医疗设备, 2022, 37(10): 18-23.
Wang ZY, Renaguli A, Wang XR, et al. Research on region segmentation algorithm of hepatic cystic echinococcosis ultrasonoscopy[J]. China Med Devices, 2022, 37(10): 18-23.
[20] 徐光柱, 钱奕凡, 王阳, 等. 基于两级分割的胎儿四腔心超声 切面质量评测[J]. 中国图象图形学报, 2023, 28(8): 2476-2490.
Xu GZ, Qian YF, Wang Y, et al. Quality assessment for fetal four-chamber ultrasound views based on two-stage segmentation[J]. J Image Graph, 2023, 28(8): 2476-2490.

基金

陕西省重点研发计划一般项目(2023-YBSF-503)。

PDF(2596 KB)

64

Accesses

0

Citation

Detail

段落导航
相关文章

/