Deep learning-based target detection algorithms have been widely used in the detection of pulmonary nodules.However,due to the small volume,variable location,and complex texture of lung tissue,the detection accuracy of pulmonary nodules still requires further improvement.In this paper,we propose an improved CenterNet algorithms for pulmonary nod-ules detection by suppressing non-target features and emphasizing nodule information.This method integrates the deep residu-al network ResNet50 and the CBAM attention module,enhancing the extraction of nodule characteristic information across different layers.Even with a deep network structure,the detection model maintains good training effectiveness.Experimental results show that our algorithm outperforms several comparative algorithms in representative evaluation metrics such as mAP,Recall and Precision,indicating that its effectiveness in pulmonary nodule detection.
万方数据
维普
