Scan-rescan reliability assessment of brain volumetric analysis across scanners and software solutions

• Published in: Scientific reports Vol. 15; no. 1; pp. 29843 - 9
• Main Authors: Bürkle, Eva, Nazzal, Ahmad, Debolski, Alexander, Ernemann, Ulrike, Lindig, Tobias, Bender, Benjamin
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 14.08.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 631/114/116; 631/114/1305; 631/114/1314; 639/705/1042; 639/705/1046; 639/705/794; 692/698/1688; 692/699/375; Adult; Artificial intelligence; Brain - anatomy & histology; Brain - diagnostic imaging; Brain research; Coefficient of variation; Confidence intervals; Female; Gray Matter - anatomy & histology; Gray Matter - diagnostic imaging; Humanities and Social Sciences; Humans; Image Processing, Computer-Assisted - methods; Magnetic Resonance Imaging - instrumentation; Magnetic Resonance Imaging - methods; Male; Medical equipment; Middle Aged; multidisciplinary; Neurodegenerative diseases; Organ Size; Reproducibility; Reproducibility of Results; Scanners; Science; Science (multidisciplinary); Software; Substantia alba; Substantia grisea; Volumetric analysis; White Matter - anatomy & histology; White Matter - diagnostic imaging
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-025-15283-3

Automated brain volumetry shows promise in improving the screening and monitoring of neurodegenerative diseases. However, the reliability of measurements across different scanners and software remains uncertain. This study assessed the reliability of gray matter, white matter, and total brain volume measurements from seven volumetry tools, using six scanners across two scanning sessions, performed within 2 h the same day, in twelve subjects. Generalised estimating equations models showed significant effects of both software and scanner on all measurements with stronger effect of software ( p  < 0.001). Percentage of coefficient of variation (CV) was calculated to measure scan-rescan reliability. Median CV across scanners of AssemblyNet and AIRAscore was less than 0.2% for grey and white matter, and 0.09% for total brain volume; while FreeSurfer, FastSurfer, syngo.via, SPM12, and Vol2Brain had a CV greater than 0.2%. In Bland-Altman analysis there was no systematic difference, but limits of agreement differed greatly between methods. Based on these findings, we recommend using the same scanner and software combination across sessions to ensure that observed changes in brain volume are reliable and clinically valuable.