Guidelines for the diagnosis of dementia have recommended the use of neuroimaging as a routine component of the initial evaluation of patients, as a supplement to clinical assessment in possible dementia cases (Knopman et al. 2001; Waldemar et al. 2007). The most recent EFNS guidelines recommend structural imaging in the evaluation of every patient affected by dementia (Filippi et al. 2012; Sorbi et al. 2012) whilst noting the difficulty of attributing clinical significance to evidence of cerebrovascular disease. Likewise national directives, such as the UK National Institute for Health and Clinical Excellence/Social Care Institute for Excellence (NICE/SCIE) guidelines (2006) state that “Structural imaging should be used in the assessment of people with suspected dementia to exclude other cerebral pathologies and to help establish the subtype diagnosis” (paragraph 1.4.3.2).

Various structural pathologies have been reported to cause dementia or cognitive impairment (Larner 2013a:166–76), including brain tumour (Case Studies 4.1 and 6.1; Ibrahim et al. 2009; Smithson and Larner 2013), subdural haematoma, and various causes of hydrocephalus, including colloid cyst (Case Study 6.2), but these have very rarely been encountered in CFC, accounting for less than 0.5 % of referrals (Larner 2013b).

Because of the ready availability of structural neuroimaging, particularly computed tomography (CT), to primary care physicians as well as to cognitive neurologists, it might be argued this investigation should be undertaken prior to referral, as is the case for “screening blood tests”. A prospective observational study of 100 consecutive patients referred to CFC (March-August 2010) collected reports of neuroimaging performed prior to and after referral (Larner 2011a). Referral source (Fig. 6.1) was predominantly from primary care (63). Of the cases referred from secondary care (37), there were equal proportions from other neurology consultants at the Walton Centre for Neurology and Neurosurgery (WCNN; 16) and from psychiatry/old age psychiatry services (16), with a handful from other sources (5), such as general physicians.

The majority of patients were not demented by DSM-IV criteria (67), although there was evidence for a neurological or neurodegenerative process causing cognitive impairment in some of these non-demented individuals (mild cognitive impairment [MCI] 10, progressive nonfluent aphasia [PNFA] 2, dementia with Lewy bodies [DLB] 2, HIV 1). One third of patients received a clinical diagnosis of dementia (33), the most frequent subtypes being AD/Down syndrome (15), frontotemporal lobar degeneration syndromes (FTLD; 8), DLB (3), and vascular dementia (2).

Patients who had undergone neuroimaging before referral to CFC were in the minority (28/100). Not unexpectedly, fewer primary care (7/63 = 11 %) than secondary care (21/37 = 57 %) referrals had undergone neuroimaging (Fig. 6.2), a statistically significant difference (χ² = 26.0, df = 1, p < 0.01). In the group of referrals from secondary care, neuroimaging frequencies by referral source were: WCNN consultant 11/16 (=69 %); psychiatry/old age psychiatry 7/16 (=44 %); and others 3/5 (=60 %).

Breakdown of the frequency of neuroimaging by patient diagnosis showed that patients ultimately receiving a diagnosis of dementia from CFC were significantly more likely to have been imaged (15/33 = 45 %) than those not demented (13/67 = 19 %; χ² = 8.14, df = 1, p < 0.01), although notably the latter group included some of those patients with underlying neurological and/or neurodegenerative disease (PNFA 2, HIV, DLB, MCI).

Following CFC consultation, all remaining patients (72) underwent neuroimaging. These data indicated that CFC was compliant with NICE/SCIE (2006) guidelines on neuroimaging. The key question though is whether this compliance makes any difference in clinical management. To a certain extent this may be a subjective value judgement, but it was the case that no revision of clinical diagnosis (e.g. not dementia revised to dementia, or a change of dementia subtype) was forthcoming as a result of CFC neuroimaging in this study (Larner 2011a), although this has happened on occasion (see Case Studies 4.1, 6.2, and 6.3). Finding potentially reversible causes of cognitive decline on neuroimaging is, sadly, exceedingly rare, the most common treatable structural causes seen in CFC being meningioma and intracranial dural arteriovenous fistula (three cases each in approximately 10 years; Wilson et al. 2010; Larner 2013b).

A second prospective observational study of neuroimaging practice prior to CFC referral was undertaken on consecutive patients over a 6-month period (September 2012-February 2013). Of 127 patients (M:F = 70:57; 55 % male; age range 19–89 years, median 61 years), the majority were not demented by DSM-IV criteria (86, =68 %), giving a dementia prevalence of 32 %, but 32 had cognitive impairment short of dementia giving a prevalence of cognitive impairment of 57 %. As before, patients who had undergone neuroimaging before referral to CFC were in the minority (51/127 = 40 %), most initiated in secondary (46) care but a few (5) in primary care. However, the proportion imaged prior to referral had increased compared to the 2010 study (40 % vs 28 %). Breakdown of the frequency of neuroimaging by patient diagnosis showed that patients receiving a diagnosis of dementia were significantly more likely to have been imaged (23/41 = 56 %) than those not demented (28/86 = 33 %; χ² = 7.38, df = 1, p < 0.01). Patients with cognitive impairment were significantly more likely to have been imaged (36/73 = 49 %) than cognitively healthy individuals (15/54 = 28 %; χ² = 6.56, df = 1, p < 0.01).

So should neuroimaging be a prerequisite for referral to memory clinics, part of the minimum dataset available to the CFC clinician prior to assessment? Probably not.

A steady trickle of patients is referred to CFC because of “cortical atrophy”, based on radiology reports of brain scans undertaken for other purposes, usually headache. (Twas ever thus: Allison (1962:257) reports a patient diagnosed with early AD because of symmetrical ventricular dilatation on air encephalography, an investigation now obsolete.) Although this label of “atrophy” may be technically correct according to neuroradiological terminology (Global Cortical Atrophy scale, grade 1 = opening of sulci), it is invariably a qualitative judgement. Moreover, experience indicates that it is seldom of clinical relevance, and serves only to generate significant (and understandable) anxiety in both patients and primary and secondary care physicians unfamiliar with the uses of neuroimaging in the assessment of cognitive problems. As with all imaging findings, clinical-radiological correlation is essential, which basically privileges the primacy of clinical assessment. A study of the diagnostic yield of CT scans done routinely in a UK memory loss clinic found “significant findings” in only 1 % (Dawe 2012). Neuroimaging provided support for, but did not alter, clinical diagnosis in a large clinico-pathological case series (Snowden et al. 2011).

With magnetic resonance imaging (MRI), the problem of incidental findings may be even more evident since they are very common (Morris et al. 2009). The presence of cerebrovascular disease in particular is problematic, as acknowledged in EFNS guidelines (Sorbi et al. 2012). Such changes seem to be interpreted by psychiatrists and old age psychiatrists as commensurate with a diagnosis of vascular dementia, despite the clarity of diagnostic criteria which require such changes to be clinically and temporally relevant to cognitive decline (Román et al. 1993; van Straaten et al. 2003; Gorelick et al. 2011).

Another incidental finding is brain calcification, particularly of the basal ganglia, said to occur in perhaps 0.5–1 % of normal CT scans, and in association with a variety of disorders (Larner et al. 2011:56–7), including Down’s syndrome (Case Study 6.4).

Focal atrophy of the medial temporal lobe (MTA) on MRI is one of the supportive criteria in recent diagnostic criteria for AD (e.g. Dubois et al. 2007; Box 6.1). MTA discriminates pathologically confirmed AD from DLB and vascular cognitive impairment (Burton et al. 2009). Visual assessment scales for MTA are available (Barkhof et al. 2011:22–3) although such assessment is not routinely undertaken at present in this centre. Structural neuroimaging is often unrewarding in suspected cases of FTLD, particularly early in the disease course. In a benign and good prognosis variant of behavioural variant frontotemporal dementia (bvFTD), or bvFTD phenocopy, neuroimaging may change little over time (Davies et al. 2006).

Other MR based modalities, such as diffusion weighted imaging (DWI), susceptibility weighted imaging (SWI), and diffusion tensor imaging (DTI) may possibly be of value in the assessment of cognitive disorders (Barkhof et al. 2011). DWI may be of particular value in suspected cases of sporadic Creutzfeldt-Jakob disease (e.g. Ali et al. 2013), SWI in cerebral amyloid angiopathy (Charidimou et al. 2012), and DTI in white matter disorders (Filley 2012). Longitudinal volumetric MR imaging measuring rates of whole-brain and hippocampal atrophy has proved a sensitive marker of neurodegeneration and may well find increasing use as a diagnostic tool and as a surrogate marker to assess treatment effects (Frisoni et al. 2010).

Of the various functional imaging modalities, 99mTechnetium hexamethylpropylene amine oxime single photon emission computed tomography (99mTc HMPAO-SPECT) is probably the most widely available, although modern diagnostic criteria for AD specify functional brain imaging with positron emission tomography (PET) in preference to SPECT (Dubois et al. 2007; Box 6.1). In vivo biomarkers for diagnostic imaging may become of greater relevance in the future, such as fluorinated PET ligands for imaging amyloid (Klunk et al. 2004; Clark et al. 2011). There is no experience of PET in CFC.

Use of HMPAO-SPECT in the assessment of cognitive problems has been standard practice in some centres, usually based on visual analysis of regional cerebral blood flow changes. Although more sophisticated analytical techniques are possible, such as statistical parametric mapping, these are not universally available. Hypoperfusion of posterior temporal and parietal regions with relative sparing of occipital blood flow is typical of AD as compared to normal controls (e.g. Montaldi et al. 1990; Talbot et al. 1998; Dougall et al. 2003), although in the visual variant of AD (posterior cortical atrophy) occipital hypoperfusion is seen. Otherwise, occipital hypoperfusion is more typical of dementia with Lewy bodies (e.g. Lobotesis et al. 2001). HMPAO-SPECT additionally has the facility for the differential diagnosis of AD and FTLD based on the frontal hypoperfusion in the latter (e.g. Talbot et al. 1998; Charpentier et al. 2000).

The most recent EFNS guidelines suggested that SPECT perfusion imaging is useful to distinguish DLB, corticobasal syndrome and Creutzfledt-Jakob disease from AD (Sorbi et al. 2012:1173).

There has been only limited experience of functional neuroimaging in CFC, perhaps surprisingly in light of the origins of isotope imaging in Liverpool (Ansell and Rotblat 1948). This has been for both logistic (no SPECT scanner on site, nearest facilities at Royal Liverpool University Hospital and Wrexham Maelor Hospital) and financial reasons. A study of the utility of HMPAO-SPECT was performed in a cohort of young cognitively-impaired patients in whom diagnostic uncertainty remained after standard clinical and neuropsychological assessment and structural brain imaging (Doran et al. 2005a). SPECT scans were visually assessed by five raters (two consultant neurologists with a specialist interest in cognitive disorders, three nuclear medicine specialists) on two occasions 6 months apart, firstly without any clinical data (“blind”), secondly with brief pertinent clinical information (“informed”). SPECT diagnoses were compared with criterion diagnoses subsequently established by the two neurologists with access to all the clinical, neuropsychological and neuroimaging data. Despite reasonable intra- and inter-rater reliability, diagnostic accuracy ranged from 32 to 58 %. SPECT scan normality or abnormality in blind and informed viewings gave respective sensitivities of 77 and 71 %, specificities of 44 and 38 %, positive predictive values of 88 and 87 %, and negative predictive values of 27 and 18 %. Calculating pairwise disease group comparisons, likelihood ratios suggested some diagnostic gain in differentiating AD from “not AD” (as also shown by Dougall et al. 2003), and in differentiating AD from FTD/focal syndromes (as also shown by Talbot et al. 1998; McNeil et al. 2007). SPECT scanning was of little help in establishing diagnoses in this (highly selected) cohort of patients, a finding which supported the conclusion of an AAN evidence-based review of SPECT imaging which concluded that SPECT could not be recommended for either the initial or the differential diagnosis of suspected dementia because it had not demonstrated superiority to clinical criteria (Knopman et al. 2001). That said, other studies in less selected cohorts than that examined here have reached different, more positive, conclusions (Talbot et al. 1998; Salmon et al. 2009).

Dopaminergic SPECT imaging (FP-CIT, DATScan), for visualisation of the dopamine transporter, may be useful to differentiate AD from DLB (Hort et al. 2010a; Sorbi et al. 2012). However, it does not pick up all DLB cases (Colloby et al. 2012). Again, logistic and financial reasons have meant that this modality has been infrequently used in CFC patients with cognitive impairment (e.g. Ali et al. 2010).

Proton magnetic resonance spectroscopy (1H-MRS) is a form of functional neuroimaging which been advocated as a possible tool for use in the evaluation and diagnosis of dementia (Kantarci 2007), based on changes in the neuronal marker N-acetyl aspartate (NAA) and in myoinositol (mI) relative to brain creatine (Cr). NAA:Cr ratios are typically decreased in disorders characterised by neuronal loss, such as AD and FTLD, and mI:Cr ratios are elevated in the presence of pathological gliosis, which may also be seen in AD and FTLD.

The utility of 1H-MRS in the diagnosis of dementia has been assessed in a highly selected population of patients attending CFC (Larner 2006). Single voxel 1H-MRS was performed on GE Signa 1.5 T Scanner (TE = 35 ms; TR = 1,500 ms), measuring NAA and mI with Cr as reference, hence generating NAA:Cr and mI:Cr ratios. Comparison of mean NAA:Cr and mI:Cr ratios in occipital and frontal voxels in demented (n = 11; AD 8, FTLD 3) and non-demented (9) patients was performed. There was a statistically significant increase in mI:Cr ratio in occipital voxels in demented patients (t = 4.60, df = 15, p < 0.001), and a trend towards reduced NAA:Cr ratio in frontal voxels in demented patients (t = 1.86, df = 13, 0.05 < p < 0.1). Acknowledging the small patient numbers and the clinical heterogeneity of cases, nonetheless this study suggested that a high occipital mI:Cr ratio may be useful in differentiating demented from nondemented patients, and a low frontal NAA:Cr may be suggestive of a diagnosis of dementia.

A subsequent study assessed the utility of 1H-MRS in the differential diagnosis of AD (n = 9) and FTLD (n = 6; bvFTD 4, semantic dementia 2) (Larner 2008b). Occipital NAA:Cr ratio was lower in AD than FTLD patients, the difference reaching statistical significance (t = 2.47, df = 13, p < 0.05), but occipital mI:Cr ratio showed no difference between the groups (t = 0.81, df = 12, 0.1 < p < 0.5). Reduced occipital NAA:Cr ratio may reflect occipital neuronal loss occurring in AD, but not in FTLD, whilst the failure of occipital mI:Cr ratio to differentiate the two conditions may reflect either the absence of occipital gliotic change in FTLD or the equality of such change in both AD and FTLD (Larner 2008b). These latter findings contrast with the previously reported utility of mI:Cr ratio in differentiating cases of dementia (increased) from non-dementia patients (Larner 2006).

In a single case, extremely high frontal mI:Cr ratio, suggesting profound gliosis, prompted a clinical diagnosis of progressive subcortical gliosis of Neumann (Neumann and Cohn 1967), in which prior structural neuroimaging had suggested a provisional diagnosis of vascular change (subcortical arteriosclerotic encephalopathy) (Larner et al. 2003). Neuroaxonal leukodystrophy also entered the differential diagnosis of this case. Case Study 6.3 provides a further example of the diagnostic utility of 1H-MRS.

Mutations in three genes have been reported to be deterministic for AD (Box 6.3): amyloid precursor protein (APP), presenilin 1 (PSEN1), and presenilin 2 (PSEN2). Of these, the most commonly affected is PSEN1 (Alzheimer Disease and Frontotemporal Dementia Mutation Database 2014). PSEN1 mutations are associated with a variable clinical phenotype which may encompass not only cognitive decline but also epileptic seizures, myoclonus, extrapyramidal features, spastic paraparesis, behavioural and psychiatric symptoms sometimes akin to those seen in bvFTD, aphasia, agnosia, and cerebellar ataxia (Larner 2011b, 2013c; Larner and Doran 2006a, b, 2009b).

Four families with PSEN1 AD have been seen in CFC, with the following mutations: p.Tyr115Cys (Doran and Larner 2006), p.Met139Val (apparently a de novo mutation; Larner and du Plessis 2003), p.Arg269Gly (with prominent behavioural and psychiatric symptoms; Doran and Larner 2004b); and p.Arg269His (with late age at onset; Fig. 6.3; Larner et al. 2007). It is recognised that there may be underascertainment of such families in the UK (Stevens et al. 2011). Concurrence of early- and late-onset AD in some families (Brickell et al. 2006) may indicate a PSEN1 mutation (Lladó et al. 2010).

Families with late-onset AD apparently transmitted as an autosomal dominant condition but without any of the known AD mutations have been reported on occasion (e.g. Jimenez-Escrig et al. 2005). One such family has been encountered in CFC, with seven affected individuals in two generations, all with onset after age 65 years as far as could be ascertained from oral family history (Hancock and Larner 2007).

AD phenotype has also been reported with some tau (MAPT) gene mutations (Sect. 6.​3.​2) including p.Arg406Trp (Rademakers et al. 2003; Tolboom et al. 2010) and the IVS10+16C>T splice site mutation (Doran et al. 2007).