Automatic MRI Volumetry Assisted Visual Assessment of the Medial Temporal Lobe in Clinical Dementia Work-Up

Abstract

Introduction:

Efficient and cost-effective diagnostic tools for supporting dementia assessment are increasingly important. We aimed to evaluate whether providing neuroradiologists with volumetric data from an automatic MRI software, NeuroQuant, enhanced the diagnostic accuracy of their visual MRI assessment.

Methods:

Two neuroradiologists assessed brain MRIs from 366 patients (mean age 67.5 years, SD 9.2, and 52% females) with subjective cognitive decline (SCD, n 79), mild cognitive impairment (MCI, n 86), or dementia (n 201). The MCI and dementia patients were further diagnosed according to an etiology of Alzheimer's disease (AD, n 217) versus non-AD (n 70). In random order the neuroradiologists visually evaluated medial temporal lobe atrophy (MTA, scale 0–4) with and without having access to the NeuroQuant report of age and sex adjusted volumetric percentiles of the hippocampus. Receiver operating characteristics (ROCs) analyses were conducted to calculate the area under the curves (AUCs) for the visual MTA, the automated NeuroQuant percentile, and the combined NeuroQuant-assisted MTA in discriminating dementia from SCD and AD from non-AD.

Results:

The AUC of the visual MTA for dementia versus SCD discrimination increased slightly but not significantly when the neuroradiologists were provided with NeuroQuant results (AUC 0.76–0.79, p 0.28). Yet, the isolated NeuroQuant evaluation reached the highest accuracy (AUC 0.85, p < 0.001), significantly better than the MTA assessment (p 0.002) and the NeuroQuant-assisted MTA (p 0.04). Only the isolated NeuroQuant assessment discriminated AD from non-AD (AUC 0.60, p 0.006).

Conclusion:

On the basis of our findings, we suggest an increased use of clinically approved automatic volumetry methods in radiological departments.