Automated 3D Segmentation of Glomeruli in Human Kidney Tissue Specimens Using 16.4 T MRI Without Contrast Agents

Aurel J. Amar, Nyoman D. Kurniawan, Luise A. Cullen-McEwen, Andrew J. Kassianos, Helen G. Healy, John F. Bertram, Shekhar S. Chandra, David C. Reutens

Abstract: Accurate and fast assessment of glomerular number in ex vivo human kidney specimens is challenging. Glomerular number is traditionally estimated using measurements on glomerular profiles, the two-dimensional samples of glomeruli seen in histological sections. Such an approach relies on complex tissue preparation protocols, specialised equipment, and staff trained to manually count glomeruli. It is labour intensive, time-consuming, and can damage the samples studied, limiting complementary analysis with other methods. In this study, we show that glomeruli can be imaged in human nephrectomy samples at ultra-high field (16.4 T) MRI without contrast agents. A three-dimensional segmentation pipeline was developed to use high-resolution MR images (30 $\mu \mathrm{m}$ isotropic resolution) to accurately estimate the number of nephrons in a sample in a fully automated manner. An independent stereological method, the physical disector/fractionator combination, was used to validate the estimates of glomerular number obtained using MRI. Statistical analyses showed a good agreement between the two methods.

External IDs:dblp:conf/isbi/AmarKCKHBCR25