A renal vascular compartment segmentation method based on dynamic contrast-enhanced images.

BACKGROUND: Kidney function assessment from renography has great potential for clinical diagnosis. Compartment models are the main analytical models in this field and the vascular compartment is the most important one, whether in the twocompartment model or three-compartment model. Currently, there are some published research studies on renal cortex segmentation. However, there are few publications introducing the methods on how to segment the vascular compartment yet. OBJECTIVE: The objective of this paper is to segment the vascular compartment automatically. METHODS: This method was tested on multi-phase scan images. A feature image reconstructed from the original images was used to segment the vascular compartment. It used the features of the time-density curve of each voxel in the contrast-enhanced images to distinguish vascular space from other areas. RESULTS:The segmentation result was evaluated by the renal glomerular filtration rate (GFR) analysis of a two-compartment model with the Patlak-Rutland technique. The dataset contained 11 kidney subjects whose GFR ranged from 19.8 ml/min to 74.9 ml/min. The results showed that the correlation between reference GFR and model derived GFR was 0.919 ( P< 0.001). CONCLUSION: Compared with segmentation performed on certain phase images, this method can avoid the problem of subjective phase selection. For a given kidney data, the proposed method can always obtain the same segmentation result automatically.