Rotation-covariant texture analysis of 4D dual-energy CT as an indicator of local pulmonary perfusion

We propose rotation-covariant texture analysis of 4D dual- energy computed tomography (DECT) as a diagnosis aid tool for acute pulmonary embolism in emergency radiology. The cornerstone of the proposed approach is to align 3D Riesz directional filters along bronchovascular structures to enable rotation-covariant comparisons of the parenchymal texture. The latter is enabled using the steerability property of Riesz filterbanks. Blockwise classification of the parenchyma from all pulmonary lobes allowed a mean area under the receiver operating characteristic (ROC) curve of 0.85, which suggests that the proposed approach can be successfully used to assist radiologists in DECT data interpretation.