Purpose To achieve intelligent detection of acute pulmonary embolism(APE)in CT pulmonary angiography based on the Res2Net attention mechanism network.Materials and Methods Retrospectively included patients with suspected of APE who underwent CT pulmonary angiography examination and were diagnosed as APE at the First Affiliated Hospital of Henan University of Chinese Medicine from February 2015 to May 2023.The dataset was randomly divided into training,validation and test set in a ratio of 7∶2∶1.The model was trained based on the Res2Net network,combined with atrous spatial pyramid pooling and attention mechanism modules,and was performed five-fold cross internal validation.Using the area under the receiver operator characteristic curve,sensitivity and specificity to assess the diagnostic performance of the model.Dice similarity coefficient,precision and intersection over union(IOU)were used to assess the segmentation performance of thrombus on the test set and plot the corresponding curves.The performance of the Res2Net attention mechanism network was compared with the classic U-Net and CE-Net model.Results A total of 303 patients with APE were included in this study.There were 212,61 and 30 cases in the training set,validation set and test set,respectively.The model's area under the curve was 0.95,sensitivity was 0.90,specificity was 1.00,Dice similarity coefficient was 0.86,precision was 0.90,Pos-IOU was 0.78 and Neg-IOU was 1.00,respectively.The parameter curves and radar chart showed that the Res2Net attention mechanism network performed better than the U-Net and CE-Net models.The visualization results of the segmentation comparison showed that the Res2Net attention mechanism network achieved higher precision in segmenting pulmonary artery thrombus.Conclusion The Res2Net attention mechanism network has good performance for detection of APE.

Purpose To achieve intelligent detection of acute pulmonary embolism(APE)in CT pulmonary angiography based on the Res2Net attention mechanism network.Materials and Methods Retrospectively included patients with suspected of APE who underwent CT pulmonary angiography examination and were diagnosed as APE at the First Affiliated Hospital of Henan University of Chinese Medicine from February 2015 to May 2023.The dataset was randomly divided into training,validation and test set in a ratio of 7∶2∶1.The model was trained based on the Res2Net network,combined with atrous spatial pyramid pooling and attention mechanism modules,and was performed five-fold cross internal validation.Using the area under the receiver operator characteristic curve,sensitivity and specificity to assess the diagnostic performance of the model.Dice similarity coefficient,precision and intersection over union(IOU)were used to assess the segmentation performance of thrombus on the test set and plot the corresponding curves.The performance of the Res2Net attention mechanism network was compared with the classic U-Net and CE-Net model.Results A total of 303 patients with APE were included in this study.There were 212,61 and 30 cases in the training set,validation set and test set,respectively.The model's area under the curve was 0.95,sensitivity was 0.90,specificity was 1.00,Dice similarity coefficient was 0.86,precision was 0.90,Pos-IOU was 0.78 and Neg-IOU was 1.00,respectively.The parameter curves and radar chart showed that the Res2Net attention mechanism network performed better than the U-Net and CE-Net models.The visualization results of the segmentation comparison showed that the Res2Net attention mechanism network achieved higher precision in segmenting pulmonary artery thrombus.Conclusion The Res2Net attention mechanism network has good performance for detection of APE.