Aims. To evaluate the diagnostic accuracy of neuroimaging (including single-photon emission computed tomography (SPECT), functional magnetic resonance imaging (fMRI), and positron emission tomography (PET)) in identifying Alzheimer’s disease (AD) compared against cognitive assessment, and/or histopathology.
Methods. We searched MEDLINE, PubMed, and Embase for studies published from 2014 onwards. We prioritised clinically representative populations - including those with mild cognitive impairment (MCI), over case-control designs which often overstate precision. Quality was assessed using Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). Subgroup analysis compared histopathology against clinical reference standards.
Results. Across 94 included studies (13 case-control, 19 cohort, 61 cross-sectional), the pooled diagnostic odds ratio (DOR) was 31.1. Sensitivity and specificity ranges varied widely across designs: case-control (66.7–100%; 53–97%), cohort (70–100%; 56–100%), and cross-sectional (20–100%; 53–100%). The broad ranges observed indicate considerable variability in diagnostic protocols. Reference standards (histopathological vs non-histopathological) significantly influenced results for DOR (p=0.0056) and sensitivity (p=0.0106). Histopathology-validated studies (n=9) demonstrated no heterogeneity in specificity (I2 of 0.0%, p=0.9143), whereas non-histopathologically validated studies showed substantial heterogeneity (I2 = 86.4%, p<0.0001).
Conclusions. The accuracy of neuroimaging in the diagnosis of Alzheimer disease varies according to imaging modality, study design, and reference standard. Reduced heterogeneity observed among histopathologically-validated studies suggests histopathology in conjunction with long term follow-up of subjects as a possible gold reference standard. Future research should utilise standardised study designs, imaging protocols, and combined clinical and histopathological diagnostic standards to ensure real-world applicability.