基于肺叶分割的支气管扩张检测与分类的研究

Bronchiectasis is a common disease, which can cause pulmonary ventilation dysfunction. Itis a long course of disease, and will cause huge social and economic burden. Deep learning is used to detect and classify bronchiectasis, so that it can process a large amount of data quickly and efficiently under the condition that the judgment accuracy is similar to that of human doctors. It can not only reduce the burden of doctors, but also provide useful reference for doctors to diagnose bronchiectasis, which has high value and significance. Image preprocessing is an important factor affecting the effect of deep learning. Lung lobe segmentation is a difficult and key point in lung CT image processing. In this study, U-Net based lung lobe segmentation method was used to process low-dose CT (LDCT) data, and then deep learning model Mask R-CNN was used to detect and classify bronchiectasis. The experimental results show that the preprocessing method of lung lobe segmentation can effectively improve the detection and classification of bronchiectasis by deep learning model. At the end of the experiment, the trained model can detect bronchiectasis in LDCT images after pulmonary lobe segmentation with an accuracy of 89.8%, the average classification accuracy is 91.0%, and the speed of recognizing an image is about 1s..