Diagnosis of dementia--automatic quantification of brain structures.

Background: The aim of the present study was to examine the usefulness of a fully automatic quantification of brain structures by means of magnetic resonance imaging (MRI) for diagnosing dementia of the Alzheimer's type (DAT).
Material and Method: MRI scans of the brains of 122 patients, referred to a memory clinic, were analysed using Neuroquant® software, which quantifies the volume of various brain structures. Clinical diagnoses were made by two doctors without knowledge of the MRI results. We performed Receiver Operating Characteristic analyses and calculated the area under the curve (AUC). A value of 1 means that all ill patients have been diagnosed as diseased and no patient has been falsely diagnosed as diseased.
Results: The mean age of the patients was 67.2 years (SD 10.5 years), 60 % were men, 63 had DAT, 24 had another type of dementia, 25 had mild cognitive impairment (MCI) and ten had subjective cognitive impairment (SCI). In the comparison between DAT patients and patients with SCI or MCI, seven of eleven volumes were significantly larger than AUC 0.5. Positive and negative likelihood ratios were less than 5 and more than 0.2, respectively, for the best limit values of the volumes. Apart from the cerebellum (AUC 0.67), none of the brain structures was significantly different from AUC 0.5 in patients with dementia conditions other than dementia Alzheimer's type.
Interpretation: MRI scans with Neuroquant analyses cannot be used alone to distinguish between persons with dementia of Alzheimer's type and persons without dementia.