Correlation Analysis of Lung Function Based on Knowledge Distillation Transformer

HAN Jiayue; LI Xitong; YAN Yuwei; GOU Shaobo; CHAI Jun

doi:10.3969/j.issn.1672-9722.2025.05.008

Abstract：Based on knowledge distillation Transformer,this paper aims to study the ability of lung high-resolution CT to dis-play the number of small airway,and to explore the application value of small airway stenosis in the diagnosis and severity grading evaluation of bronchial asthma.50 patients with bronchial asthma and control group is 15 in this study,and each patient had a high-resolution CT scan and lung function related examination.Then,observe and record the number of narrow small airway and the proportion of narrow small airway in each segment of double lung and collect the relevant parameter data of pulmonary function test.Specifically,this paper leverages knowledge distillation Transformer diagnosis network to analyze the relationship of the num-ber of narrow small airway and pulmonary function.The number and proportion of small airway with double pulmonary stenosis in pa-tients with bronchial asthma are significantly correlated with pulmonary function classification(P＜0.001).There are significant dif-ferences in the number of narrow small airway between patients with roughly normal pulmonary function and those with moderate,se-vere or very severe pulmonary dysfunction(P<0.05).There exists significant difference in the number and the proportion of narrow small airway between patients with mild pulmonary dysfunction and those with severe or very severe pulmonary dysfunction(P<0.05).The proposed knowledge distillation Transformer can show the number of small airway and narrow small airway,and provide image basis for the evaluation of bronchial asthma.The number and proportion of narrow small airway can be utilized to evaluate the severity of pulmonary function in asthmatic patients.