Lesion feature enhancement and boundary-aware fusion for pulmonary embolism recognition on CT images

The article presents an automatic method for recognizing pulmonary embolism on CT images that combines lesion feature enhancement and boundary-aware structural information fusion. The method uses a Pulmonary Embolus Feature Enhancement Module and a Vascular Boundary Aware Fusion Module to improve lesion representation and boundary structure modeling. When tested on 523 CT cases from a single center, the method achieved accuracy of 0.956, precision of 0.961, recall of 0.951, and AUC of 0.971. In external validation on FUMPE and RSNA datasets, it achieved accuracy of 0.779 and 0.672 respectively. The results indicate the method's potential for computer-aided pulmonary embolism screening.