16 MRI and the Differential Diagnosis of Dementia António J. Bastos-Leite and Philip Scheltens Introduction —————————————— —————————————— Structural neuroimaging with either computed tomography (CT) or magnetic resonance (MR) imaging is recommended for the initial evaluation of patients with dementia (Knopman et al., 2001), and is being increasingly used to support the clinical diag- nosis beyond the traditional exclusionary approach (Scheltens et al., 2002). Alzheimer’s disease (AD) is the most common cause of dementia, with prevalence rates higher than 40% at the age of 85 (Evans et al., 1989), but a large proportion of patients with dementia have a combination of different types of pathology in the brain (Esiri et al., 1999; Heyman et al., 1998; Holmes et al., 1999; Hulette et al., 1997; Ince et al., 1991; Kalaria and Ballard, 1999; Lim et al., 1999; MRC CFAS, 2001; Snowdon et al., 1997). Vascular dementia (VaD), dementia with Lewy bodies (DLB), Parkinson disease with dementia (PDD), frontotemporal lobar degeneration (FTLD), and some rare atypical parkinsonian syndromes are the other most well-known causes of dementia next to AD. Although there are established clinical criteria for the diagnosis of diseases causing dementia (McKeith et al., 1996; McKhann et al., 1984; Neary et al., 1998; Roman et al., 1993), the definite diag- nosis was always believed to be histopathological. However, there are considerable discrepancies between different postmortem pathological criteria that render postmortem diagnosis less a “gold standard” than previously assumed (Nagy et al., 1998; Polvikoski et al., 2001). Clinical information is always needed for a correct classification and should always lead the diagnostic process. Next to this, neuroimaging plays a very useful role by helping to depict underlying pathology (see Chapter 3 for a discussion of the Neuropathology of Aging). Currently, brain atrophy and cerebrovascular disease are the two most important characteristics in the evaluation of dementia by means of structural neuroimaging. MR is currently the preferred imaging modality for dementia. T1-weighted images (T1-WI) are needed to detect brain atrophy, especially in the medial temporal lobe, for which coronal high-resolu- tion images are required. For the detection of cerebrovascular pathology, the MR imaging protocol should also include axial T2-weighted images (T2-WI), axial fluid-attenuated inversion recovery (FLAIR) or proton density-weighted images (PD-WI), and axial gradient-echo T2*-weighted images (T2*-WI). T2*-WI are particularly useful in detecting microbleeds (Bastos Leite, Scheltens, and Barkhof, 2004a). Alzheimer’s Disease —————————————— —————————————— Because neuropathological changes underlying late-onset AD first occur in the medial temporal lobe (Braak and Braak, 1991), structural neuroimaging in AD has been focused on the detection of medial temporal lobe atrophy (MTA), particularly of the hippocampus, parahippocampal gyrus (including the entorhinal cortex), and amygdala. MR and CT are indeed sensi- tive to MTA in AD (de Leon et al., 1989; Kido et al., 1989; Seab et al., 1988), correlating with Alzheimer pathology at post- mortem (Bobinski et al., 2000; Davis et al., 1995). MTA can be assessed by using visual rating scales (Scheltens et al., 1992a), linear measurements of temporal lobe structures (Bastos-Leite et al., 2006a; Frisoni et al., 1996; Jobst et al., 1992), and volu- metry of the hippocampus (Jack et al., 1992). e most well- known visual rating scale for MTA (Scheltens et al., 1992a) is based on the evaluation of the choroidal fissure width, the temporal horn width, and the hippocampal height (Table 16.1) using coronal high resolution T1-WI perpendicular to the long axis of the temporal lobe (Figure 16.1). It is easily applicable in clinical practice, but slightly observer-dependent (Scheltens et al., 1995). MTA scores of 2 or less are not frequently associated with dementia (Barkhof et al., 2007). e fimbriosubicular distance is a linear measurement that enables evaluation of the hippocampal sulcus width (Figure 16.2), and complements the visual rating scale for MTA (Bastos-Leite et al., 2006a). Linear measurements of the temporal horn width can be used in routine clinical settings, and have the advantage of being applicable both to CT and MR (Frisoni et al., 1996, 2002). Volumetric analyses are time-consuming and therefore not well-suited for clinical practice (Wahlund et al., 1999). Besides the existence of MTA, the most important struc- tural imaging feature of AD is the progression of hippo- campal atrophy. Jack et al. (1998) have found a yearly decline in hippocampal volume that is approximately 2.5 times greater in patients with AD than in normal-aged subjects, and a rela- tion that exists between memory loss and hippocampal damage across the spectrum from normal aging to dementia (Petersen et al., 2000). Visual rating scales are relatively insensitive to changes of hippocampal volume over time. In addition, the corresponding neuroanatomical changes may be too mild, diffuse, or topographically complex to be detected even by manually traced measurements of regions of interest. New serial volumetric imaging techniques developed in the past few years 261