Pelvic nerves are required for normal baseline functions of the pelvic organs and provide a target for therapeutic neuromodulation (neurostimulation). Unlike somatic nerves (e.g. sciatic), pelvic and other visceral nerves mostly comprise small diameter myelinated or unmyelinated axons (i.e. afferent Adelta- and C-fibers, or efferent fibers of autonomic spinal or ganglionic visceral motor neurons). In recent studies we found quantitative measurements of primary morphological parameters that define action potential conduction in visceral nerves is limited relative to somatic nerves. To address this, we studied 3D ultrastructural properties of myelinated axons in the rat pelvic nerve. First, a 93 micrometer segment of a rat pelvic nerve fascicle was imaged using scanning block-face electron microscopy. We then developed a custom pipeline (3D-VAx) to analyze the resulting large 3D image and obtain morphological parameters from all myelinated axons, including measurements of nodes of Ranvier and paranode regions. Myelin, myelinated and unmyelinated axons were auto-segmented with a novel machine learning approach trained with a small number of human annotations. To validate model accuracy, we compared 3D spatial morphometric characteristics measured from 3D-VAx segmentation results and human annotation of the node region of selected myelinated axons. This demonstrated a significant advance over currently used U-Net models. Additional capabilities of 3D-VAx include node auto-detection, total count estimation and computation of the centroid B-spline and perpendicular cross section in tortuous axons. Development of this pipeline provided new insights into the structure of the pelvic nerve and established a powerful approach for analyzing peripheral nerves in 3D.