CT model combining U-Net segmentation and deep learning distinguishes tuberculous from malignant pleural effusion

A CT deep-learning model combining U-Net segmentation and clinical features distinguished tuberculous from malignant pleural effusion with test-set area under the receiver operating characteristic curve (AUC) 0.934 (95% CI 0.8733-0.9955), sensitivity 0.875, specificity 0.900.

• The study included 281 patients with etiologically or pathologically confirmed pleural effusion, 143 with tuberculous and 138 with malignant effusion.
• The U-Net segmentation model achieved Dice coefficients of 0.873 in training and 0.862 in testing.
• The comprehensive model outperformed clinical, radiomics, and two-dimensional deep learning models, which had test-set AUCs of 0.767, 0.841, and 0.776, respectively.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.