Segmentation in 3D MRI After Injection of Cationic Nanoparticles

11 The glomeruli of the kidney perform the key role of blood filtration and the number of glomeruli in a 12 kidney is correlated with susceptibility to chronic kidney disease and chronic cardiovascular disease, 13 driving interest in new technology such as magnetic resonance imaging (MRI) to measure kidney 14 morphology in vivo. Magnetic nanoparticles of cationic ferritin have been used as MRI contrast agents to 15 target, image, and count individual glomeruli in 3D in the whole kidney with gradient-echo MRI. 16 Accumulated nanoparticles create punctate spots in the MR images associated with the glomeruli. 17 However, there currently lack of computationally efficient techniques to perform fast, reliable and 18 accurate counts of glomeruli in a MR image. To fill in this gap, we propose a three-phased pipeline 19 termed Hessian based multi-Features Clustering (HmFC). The first phase uses the convexity property 20 from Hessian matrix to highlight small regions in a whole image, with each region being candidate 21 glomerulus. Six features are then derived from imaging intensity, domain knowledge and geometric 22 information for the candidate glomeruli. A variational Bayesian Gaussian Mixture model is implemented 23 to identify the true glomeruli. We first use 2D histological images from literature test the performance of 24 HmFC against watershed and graph-cut methods. Four rat 3D MR images are studied to test the 25 applicability of HmFC for segmenting renal glomeruli. 26