PURPOSE: To present a novel postprocessing technique for artery/vein separation and background suppression from contrast-enhanced time-resolved magnetic resonance angiography datasets in order to improve the diagnosis of vessel pathologies and arteriovenous fistulas. MATERIALS AND METHODS: Ten normal, two pathologic datasets of the brain, and one hand angiography dataset were postprocessed. Cross-correlation maps between the signal time course of every voxel in the dataset and selected arterial and venous references regions of interest (ROIs) were obtained; these maps were subsequently nonlinearly transformed to obtain two indices representing the likelihood of a voxel belonging to a vessel category. Red-green-blue (RGB) color encoding was utilized to depict synthetic arteriogram and venogram images in a single diagnostically meaningful image. RESULTS: The technique enabled correct visual separation of vessels on various datasets, as evaluated by two expert neuroradiologists, and also highlighted characteristics of flow in arteriovenous fistulas. A quantitative comparison with existing techniques showed better separation performance on 3 out of 10 normal datasets and higher stability to acquisition characteristics and contrast agent bolus dispersion. CONCLUSION: This method can be helpful in the diagnosis of vascular diseases in subjects where bolus dispersion makes it difficult to discriminate between arteries and veins with standard methods (subtraction or correlation analysis). Copyright (c) 2008 Wiley-Liss, Inc.
PURPOSE: To present a novel postprocessing technique for artery/vein separation and background suppression from contrast-enhanced time-resolved magnetic resonance angiography datasets in order to improve the diagnosis of vessel pathologies and arteriovenous fistulas. MATERIALS AND METHODS: Ten normal, two pathologic datasets of the brain, and one hand angiography dataset were postprocessed. Cross-correlation maps between the signal time course of every voxel in the dataset and selected arterial and venous references regions of interest (ROIs) were obtained; these maps were subsequently nonlinearly transformed to obtain two indices representing the likelihood of a voxel belonging to a vessel category. Red-green-blue (RGB) color encoding was utilized to depict synthetic arteriogram and venogram images in a single diagnostically meaningful image. RESULTS: The technique enabled correct visual separation of vessels on various datasets, as evaluated by two expert neuroradiologists, and also highlighted characteristics of flow in arteriovenous fistulas. A quantitative comparison with existing techniques showed better separation performance on 3 out of 10 normal datasets and higher stability to acquisition characteristics and contrast agent bolus dispersion. CONCLUSION: This method can be helpful in the diagnosis of vascular diseases in subjects where bolus dispersion makes it difficult to discriminate between arteries and veins with standard methods (subtraction or correlation analysis). Copyright (c) 2008 Wiley-Liss, Inc.