Exploratory Non-Inferiority Validation of an East Asian-Based MRI Brain Volumetric Software

Purpose: Accurate measurement of brain volume using structural magnetic resonance imaging (MRI) is crucial for diagnosing and monitoring of Alzheimer’s disease. This study aimed to clinically validate a new MRI-based brain volumetric software (the study soft-ware) against a widely used and validated reference software. Materials and Methods: Patients with cognitive impairment who underwent three-di-mensional structural brain MRI between January and December 2021 were retrospec-tively included. The software’s performance in detecting significant medial temporal lobe atrophy (MTA), along with the correlation between Scheltens’ MTA scores and hippo-campal volumes, was assessed using a non-inferiority test and Spearman’s correlation analysis, respectively. MTA scores were determined by consensus among three neurora-diologists. Results: A total of 271 patients (mean age 79 ± 7 years [standard deviation], 178 wom-en) were included. The hippocampal volumes calculated by the two software programs were not equivalent. In detecting significant MTA, the study software demonstrated a non-inferior area under the receiver operating characteristic curve (AUC) (0.78 [95% con-fidence interval (CI): 0.73, 0.84]), accuracy (0.77 [95% CI: 0.72, 0.82]), and specificity (0.89 [95% CI: 0.84, 0.93]) compared to the reference software’s AUC (0.71 [95% CI: 0.65, 0.78]), accuracy (0.72 [95% CI: 0.67, 0.78]), and specificity (0.80 [95% CI: 0.73, 0.85]) (p < 0.05). However, non-inferiority in sensitivity was not established between the study software (0.49 [95% CI: 0.38, 0.61]) and reference software (0.55 [95% CI: 0.44, 0.66]). The correlation between MTA scores and hippocampal volume was significantly stronger in the study software (-0.57 [95% CI:-0.64,-0.48]) than in the reference software (-0.30 [95% CI:-0.41,-0.19]) (p < 0.05). Conclusion: The new East Asian cohort-based MRI brain volumetric software demonstrated non-inferiority to the reference software in detecting significant MTA. Further-more, the new study software showed a greater association with the neuroradiologists’ MTA scores than the reference software.