Structural MRI-based Computer-aided Diagnosis Models for Alzheimer Disease: Insights into Misclassifications and Diagnostic Limitations

Abstract

Purpose To examine common patterns among different computer-aided diagnosis (CAD) models for Alzheimer disease (AD) using structural MRI data and to characterize the clinical and imaging features associated with their misclassifications. Materials and Methods This retrospective study used 3258 baseline structural MRI scans from five multisite datasets and two multidisease datasets collected between September 2005 and December 2019. The 3D Nested Hierarchical Transformer (3DNesT) model and other CAD techniques were used for AD classification using 10-fold cross-validation and cross-dataset validation. Subgroup analysis of CAD-misclassified individuals compared clinical and neuroimaging biomarkers using independent t tests with Bonferroni correction. Results This study included 1391 patients with AD (mean age, 72.1 years ± 9.2 [SD]; 757 female), 205 with other neurodegenerative diseases (mean age, 64.9 years ± 9.9; 117 male), and 1662 healthy controls (mean age, 70.6 years ± 7.6; 935 female). The 3DNesT model achieved 90.0% ± 2.3 cross-validation accuracy and 82.2%, 90.1%, and 91.6% accuracy in three external datasets. Further analysis suggested that the false-negative subgroup (n = 223) exhibited minimal atrophy and better cognitive performance on the Mini-Mental State Examination (MMSE) than the true-positive subgroup (MMSE score in false-negative subgroup, 21.4 ± 4.4; true-positive subgroup, 19.7 ± 5.7; P value family-wise error [P_FWE] < .001), despite displaying similar levels of amyloid β (false-negative subgroup, 705.9 pg/mL; true-positive subgroup, 665.7 pg/mL; P_FWE = .99) and tau (false-negative subgroup, 352.4 pg/mL; true-positive subgroup, 371.0 pg/mL; P_FWE = .99) burden. Conclusion A subgroup of patients with false-negative classification for Alzheimer disease exhibited atypical structural MRI patterns and clinical measures, fundamentally limiting the diagnostic performance of CAD models based solely on structural MRI. Keywords: MR Imaging, Dementia, Computer Applications-3D, Alzheimer's Disease, Computer-aided Diagnosis, Misclassification, Atypical AD Supplemental material is available for this article. © RSNA, 2025 See also commentary by Nasrallah in this issue.

Keywords: Alzheimer’s Disease; Atypical AD; Computer Applications–3D; Computer-aided Diagnosis; Dementia; MR Imaging; Misclassification.