Kidney segmentation from a single prior shape in MRI

This paper reports a novel approach to 3D kidney segmentation from a single prior shape in magnetic resonance imaging (MRI) datasets. The proposed method is based on a hierarchic surface deformation algorithm, to generate a pre-personalized model, followed by an anamorphing segmentation algorithm, to extract the kidney capsule. Accuracy and precision are assessed by comparing our method over 20 kidney reconstructions segmented manually by 3 different observers on native MRI images. The experimental results show a volumetric overlap error of 6.39±2.47%, a relative volume difference of 1.87±1.39%, an average symmetric surface distance of 0.80±0.23mm, a root mean squared symmetric distance of 1.03±0.33mm and a maximum symmetric surface distance of 4.18±3.45mm. With our method, the capsules of both kidneys are segment in less than 40 seconds.