Identification of Transcranial Doppler Blood Flow Parameters for
Dizziness Based on L2,1 Norm of Machine Learning
PENG Jing1a, ZOU Yihuai1a, SU Jiaming1b, WU Kang1a, SONG Fan2, CHEN Xing1c
1. a. Department of Neurology; b. Department of Endocrinology; c. Department of Brain Function Examination,
Dongzhimen Hospital, Beijing University of Chinese Medicine, Beijing 100700, China;
2. School of Biological and Medical Engineering, Beihang University, Beijing 100191, China
Abstract：Objective To explore the diagnostic value of transcranial Doppler (TCD) blood flow characteristic parameters in dizziness
by using TCD technology based on L2,1 norm of machine learning method. Methods A total of 41 patients with dizziness and 21
healthy subjects from Dongzhimen Hospital, Beijing University of Chinese Medicine were selected as subjects. The blood flow
signals of bilateral middle cerebral artery, terminal segment of bilateral internal carotid artery, siphon segment of bilateral internal
carotid artery, bilateral anterior cerebral artery, bilateral posterior cerebral artery, bilateral vertebral artery, and proximal and distal
basilar artery were recorded by TCD diagnostic instrument. At the same time, the peak systolic velocity, end-diastolic velocity,
average velocity, pulse index and resistance index of the above 14 vascular sites were measured, and the L2,1 norm method was used
to take the above parameters and age as features, and input them into the machine learning classifier after feature selection to classify
and predict patients with dizziness. Overall accuracy and area under ROC curve (AUC) were used as the evaluation indexes of the
model. Results Important features such as right vertebral artery end diastolic flow velocity, right middle cerebral artery peak systolic
flow velocity, right vertebral artery peak systolic flow velocity, right vertebral artery resistance index and right middle cerebral artery
end diastolic flow velocity contributed the classification of dizziness most, the model based on ensemble learning algorithm was
the best, with the AUC value was 0.906. Conclusion Ensemble learning algorithm may be more suitable for the classification of
TCD blood flow features in patients with dizziness. The cerebral hemodynamic changes in patients with dizziness are mainly in the
vertebrobasilar artery system, and may also effect the middle cerebral artery and other internal carotid artery systems.
彭景1a，邹忆怀1a，宿家铭1b，吴康1a，宋凡2，陈星1c. 基于L2,1范数机器学习方法的头晕经颅多普勒超声血流特征参数识别[J]. 中国医疗设备, 2022, 37(10): 24-28.
PENG Jing1a, ZOU Yihuai1a, SU Jiaming1b, WU Kang1a, SONG Fan2, CHEN Xing1c. Identification of Transcranial Doppler Blood Flow Parameters for
Dizziness Based on L2,1 Norm of Machine Learning. China Medical Devices, 2022, 37(10): 24-28.
