Population-Scale Segmentation of Penile Tissue in DIXON MRI using Deep Learning for Quantitative Phenotyping in Male Reproductive Health

2026-07-02

AI Summary

Current methods struggle to achieve automated segmentation of penile tissues at population scale, hindering standardized quantitative phenotyping in male reproductive health research. This study addresses this gap by developing a fully automated segmentation model based on the 3D nnU-Net architecture applied to multi-contrast DIXON MRI, enabling high-precision delineation and volumetric quantification of the entire penis—including both internal and external structures—for the first time. Trained on expert-annotated data from 145 participants, the model achieved a cross-validated Dice score of 0.90 and 0.92 on an independent, double-blinded test set of 24 cases. Deployed across 34,412 participants in the UK Biobank, the model demonstrates excellent longitudinal reproducibility (intraclass correlation coefficient r = 0.87), offering a scalable tool for large-scale imaging-based phenotypic analysis of male reproductive health.

Abstract

Penile measurement is clinically relevant across male reproductive and urogenital health, including conditions such as micropenis, congenital and endocrine disorders, and sexual or urinary dysfunction. However, quantitative assessment of penile size has relied mainly on external length or circumference measurements, which are difficult to standardize, sensitive to measurement conditions, and unable to capture the internal portion of the penis. MRI enables volumetric assessment of the whole penis in vivo, but automated segmentation has not previously been established at population scale. Automated whole-organ volumetry would enable high-throughput phenotyping for multi-omics and clinical studies of male reproductive disease. Here, we present a deep learning framework for whole-penis segmentation in multi-channel DIXON MRI. Using a newly curated expert-annotated training dataset ($n = 145$ subjects; $13,050$ annotated slices) and a double-annotated independent test benchmark ($n = 24$ subjects; $2,160$ double-annotated slices), we optimized a 3D nnU-Net architecture. The model achieved a 5-fold cross-validation Dice score of $0.90$ and performed at observer-level accuracy on the independent test set (Dice: $0.92$; Hausdorff distance: $3.58$). We deployed the model in $34,412$ UK Biobank participants, enabling automated quantification of total penile tissue, including both external and internal components. Longitudinal evaluation in 2,282 men demonstrated high inter-session reproducibility ($r = 0.87$). This framework establishes a reproducible and population-scalable method for MRI-based assessment of penile anatomy and provides an open technical resource for future studies in urological imaging and male reproductive health. The trained model weights will be publicly released.

Problem

• penile segmentation
• quantitative phenotyping
• male reproductive health
• DIXON MRI
• population-scale

Innovation

• deep learning
• penile segmentation
• DIXON MRI
• population-scale phenotyping
• nnU-Net