Agreement between volBrain and mdbrain AI software for thalamic and white matter lesion volumetry in multiple sclerosis

Rodrigo I. Brito-Delgado, Department of Magnetic Resonance, Hospital Angeles Lomas, Huixquilucan, State of Mexico, Mexico;
Eduardo Torres-Olivas, Department of Magnetic Resonance, Hospital Angeles Lomas, Huixquilucan, State of Mexico, Mexico;
Arturo Hernandez-Medina, Department of Magnetic Resonance, Hospital Angeles Lomas, Huixquilucan, State of Mexico; Biomedical Engineering Area, Faculty of Engineering, Universidad Autonoma de Queretaro, Queretaro, Qro. Mexico;

Introduction: Artificial intelligence (AI)-based software platforms utilizing magnetic resonance imaging (MRI) have been developed to quantify brain structures and white matter lesions (WML) in patients with multiple sclerosis (MS). This study evaluated the agreement between volBrain and mdbrain software for thalamic and WML volumetry in patients with MS. Material and methods: This retrospective cross-sectional study included patients with MS. MRI examinations were performed on 1.5T or 3.0T scanners. Thalamic and WML volumes obtained with volBrain software were compared with those generated by mdbrain software. Statistical analysis included Student’s paired t-test or the Wilcoxon signed-rank test, correlation analysis, and the Bland-Altman agreement assessment. Results: In 31 MS patients, the mean thalamic volume with volBrain was 14.48 ± 1.88 mL; the mean with mdbrain was 14.29 ± 1.94 mL. volBrain and mdbrain showed excellent thalamic volumetry agreement, with a bias of -0.194 mL and narrow limits of agreement (-1.199, 0.810) without proportional bias. WML volumetry exhibited significant disagreement between volBrain and mdbrain, with median volumes of 3.32 mL (IQR 5.08) and 4.48 mL (IQR 6.25), respectively. WML volumetry demonstrated minimal bias (0.062 mL) but wider limits of agreement (-5.588, 5.713) and significant proportional bias (p < 0.001). Conclusion: Agreement between volBrain and mdbrain for thalamic volumetry using MRI in patients with MS was excellent. However, proportional bias was observed in WML volumetry, particularly with high WML volume. VolBrain showed greater disagreement with mdbrain as lesion volume increased, suggesting that AI software may not be interchangeable for evaluating patients with MS.

Keywords: Multiple sclerosis. Thalamus. White matter lesion. mdbrain. volBrain. Artificial intelligence.