Related Experiment Video
Updated: Sep 14, 2025
Analysis of Lymph Node Volume by Ultra-High-Frequency Ultrasound Imaging in the Braf/Pten Genetically Engineered Mouse Model of Melanoma
Published on: September 8, 2021
An Automatic Arrival Time Parametric Imaging Method for the Classification of Blood Perfusion Direction of Lymph
Hui Yu1, Qingsong Wang1, Yiming Gao2
1Department of Biomedical Engineering, Tianjin University School of Medicine, Tianjin, China; State Key Laboratory of Advanced Medical Materials & Devices, Tianjin University School of Medicine, Tianjin, China.
Objective:
Arrival time parametric imaging (AT-PI) of contrast-enhanced ultrasound (CEUS) videos plays a crucial role in the clinical diagnosis of lymph node disease. However, imaging quality depends on manually set parameters. To improve the automation and objectivity of the CEUS diagnostic process, we propose an automatic AT-PI method and a classifier for identifying blood perfusion direction (BPD).
Methods:
A total of 120 CEUS videos were collected from 120 patients (140 nodes) with lymph node disease. The important principle for admission was that the BPD conclusions by three doctors were the same. Three tasks were devised: i) a localization task to locate lymph nodes was realized by a YOLOv7 model followed by a Kalman filter; ii) the second task was automatic AT-PI, which utilized a Likely Rectified Linear Unit (ReLU) method to fit the time-intensity curve of CEUS videos; iii) in the third task, a ResNet50 model was employed to divide the lymph nodes into two BPDs-centripetal and centrifugal.
Results:
The first task reached a mean average precision (mAP) of 57.3% in the test set with 10 videos. For the second task, three doctors were required to provide BPD conclusions based on our AT-PI images, while the diagnostic results and the ground truth showed a Kappa of 0.916 in 140 nodes. The classification task also achieved an average F1-score of 95.8% in the fivefold cross-validation experiment.
Conclusion:
The proposed automatic AT-PI method and classifier of BPD with high-quality validation results show great clinical potential for improving the efficiency and objectivity of CEUS examinations.
