Field of the Invention

The present invention relates to a method, and particularly to a method of determining a likelihood that a patient has a neurological condition.

Background

The Fronto-Temporal Lobar degeneration (FTLD) spectrum covers several clinically and neuropathological distinct subtypes (Bang et al., 2015). Behavioural variant frontotemporal dementia (bvFTD) is characterised by insidious onset and progressive deterioration with disinhibition, apathy, lack of empathy, compulsions, hyperorality, and impairment of executive dysfunction (Rascovsky et al. 2007). There are also patients who exhibit similar symptoms, but do not suffer from a neurodegenerative condition (Kipps et al., 2008), characterised by preservation of functional ability, normal imaging, and absence of a genetic abnormality (Rohrer 2011). Recently, higher levels of socially inappropriate behaviour and criminality in bvFTD as compared to Alzheimer’s disease (Liljegren et al., 2019) has been recognised, due to a dissociation affecting the capacity for an appropriate evaluation of actions and their consequences whilst retaining a cognitive grasp (Mendez 2010; Sfera et al, 2014).

Patterns of clinical and behavioural impairment vary within the bvFTD subtype (Josephs et al., 2009; Rohrer et al., 2009), as do patterns in brain atrophy identified by magnetic resonance imaging (MRI) (Ranasinghe et al., 2016; Whitwell et al., 2009; Whitwell et al., 2013). It has been proposed that the functional changes in the brain linked to different patterns of disturbance in brain connectivity, collectively referred to as network vulnerability, might play an important role, an idea supported by electroencephalographic (de Haan et al., 2009) and functional MRI studies (Zhou et al., 2010). Changes in the so-called salience network, responsible for processing of behaviourally salient stimuli, is a well-documented neuroimaging feature of bvFTD (Seeley et al., 2007). However, alternations in the functional connectivity in the default mode network has also been reported (Irish et al. 2012). Furthermore, there is increasing evidence that multiple cognitive control networks and their overlapping sub-systems are impaired in bvFTD (Ranasinghe et al., 2016). Summary

Accordingly, in a first aspect, embodiments of the present invention provide a method of determining an indication as to whether a patient has a neurocognitive disease, the method comprising: obtaining structural neurological data from a patient, said structural neurological data being indicative of a physical structure of one or more parts of the following territories of the patient’s brain: limbic lobe, frontal lobe, and sub-cortical grey nuclei; calculating a volume of the or each part of each territory; and determining, from the calculated volume of the or each part of each territory, an indication that the patient has a neurocognitive disease.

Surprisingly, it has been found that focusing on the territories identified above yields an improved diagnosis of neurocognitive diseases.

The method may have any one, or any combination insofar as they are compatible, of the following optional features.

By territory, it may be meant a group of brain regions as parcellated by a brain atlas (preferably the Desikan-Killiany atlas). By part, it may be meant specific sets of parcellated region of the brain. The parts of the limbic lobe forwhich volume is calculated may include any one or more of, and preferably all of: the anterior cingulate gyrus, hippocampal and parahippocampal gyri, insula, and the superior temporal gyrus.

The parts of the sub-cortical grey nuclei for which volume is calculated may include any one or more of, and preferably all of: the amygdala, the caudate nucleus, the pallidum, and the thalamus.

The data may be further indicative of a physical structure of one or more parts of one or more of the following territories of the patient’s brain: central region, and temporal lobe. The part of the central region forwhich volume is calculated may include the precentral gyrus. The parts of the frontal lobe forwhich volume is calculated may include any one or more of, and preferably all of: the orbital surfaces of the frontal cortex and the medial surfaces areas of the frontal cortex. The part of the temporal lobe forwhich volume is calculated may include the middle temporal gyrus. The neurocognitive disease may be one of Alzheimer’s disease and behavioural variant fronto-temporal dementia.

The structural neurological data may be magnetic resonance imaging data.

Determining the likelihood that the patient has a neurocognitive disease may be performed by a trained single-layered artificial neural network. The artificial neural network may be a single-layered perceptron neural network. The trained single-layered artificial neural network may take as an input a volume of the or each part of each territory, and provide as an output a classification indicative of whether the patient has or does not have a neurocognitive disease. The input may be one or more regional volumes of the or each part of each territory.

Calculating the volume in the or each part of each territory within the brain may be performed using voxel-based-morphometry. The calculation of the volume may be performed voxel-wise or based on an assumed parcellation of the brain. By voxel-wise calculation, it may be meant that the calculation of volume is performed without an assumption as to the labelling of particular brain volumes and its boundaries, for example by the pair-wise comparisons of voxels of a control, reference group’s brain to the voxels of the patient’s group brain. A calculation of a degree of atrophy may be performed assuming specific labelling of particular brain anatomies (for example, after positive identification of the hippocampus, given the brain atlas, by the comparison of reference volumes for the hippocampus to the hippocampus volumes of the patients, at the group level). Said another way, the volume of the or each part of each territory may be compared voxel-by-voxel or by aggregation of voxels (labelled regions), to an expected value of a normal control sample or reference volume (the reference volume being calculated from a healthy or control group) to determine the degree of atrophy forthat part. The calculation of the volume or the degree of atrophy for the or each part of each territory may be performed using an output of the FreeSurfer software package.

The method may be a computer-implemented method.

In a second aspect, embodiments of the present invention provide a system for determining an indication as to whether a patient has a neurocognitive disease, the system comprising: data obtaining means, configured to obtain structural neurological data from a patient, said structural neurological data being indicative of a physical structure of one or more parts of the following territories of the patient’s brain: limbic lobe, frontal lobe, and subcortical grey nuclei; calculation means, configured calculate a volume of the or each part of each territory; and determination means, configured to determine, from the calculated volume of the or each part of each territory, an indication that the patient has a neurocognitive disease.

Surprisingly, it has been found that focusing on the territories identified above yields an improved diagnosis of neurocognitive diseases. Any of the methods or systems used herein for determining an indication as to whether a patient or subject (the terms are used interchangeable) has a neurocognitive disease may be used in a method or system for diagnosing that disease in the patient or subject, or assessing the likelihood or risk that the patient or subject has a neurocognitive disease.

The structural neurological data-based measures described herein may optionally be combined with (prior to, following, or in conjunction with) other criteria which are relevant to the disease or diseases in question. Such other criteria may include:

• the age of the subject;

• psychometric outcome measures;

• in relation to bvFTD, meeting criteria for probable bvFTD according to the International Consensus Criteria for bvFTD (Rascovsky et al., 2011 , Sensitivity of revised diagnostic criteria for the behavioural variant of frontotemporal dementia. Brain 134:2456-2477); see in particular Table 3 therein;

• in relation to bvFTD, a Centrally rated frontotemporal atrophy score e.g. of 2 or greater (Kipps et al., 2007, Clinical significance of lobar atrophy in frontotemporal dementia: application of an MRI visual rating scale. Dementia Geriat. Cognit. Disord. 23:334-342);

Psychometric outcome measures for use in the methods may be conventional ones, as accepted by appropriate regulatory bodies.

For AD, the Alzheimer’s Disease Assessment Scale - cognitive subscale [ADAS-cog] is preferred (Rosen WG, Mohs RC, Davis KL. A new rating scale for Alzheimer's disease. Am J Psychiatry. 1984 Nov;141 (11):1356-64). Another standardised test is the Mini-Mental State Examination [MMSE] which was proposed as a simple and quickly administered method for grading cognitive function (Folstein MF, Folstein SE & McHugh PR. ‘Mini-mental state’. A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research 1975 12 189— 198.). The MMSE is the most widely used cognitive screening instrument for the detection of cognitive dysfunction due to dementia in geriatric and psychiatric patients (Tombaugh TN & McIntyre NJ. The mini-mental state examination: a comprehensive review. Journal of the American Geriatric Society 1992 40 922-935). The MMSE evaluates orientation, memory, attention and language functions.

Further measures which may also be utilised include (1) Addenbrooke's Cognitive Examination - revised (ACE-R, Mioshi et a/ 2006) and (2) Functional Activities Questionnaire (FAQ, Pfeffer et al 1982).

The methods and systems described herein thus have utility in making or assisting diagnosis, and in turn in selecting subjects for treatment with appropriate prophylactic or therapeutic measures, or admission into a clinical trial of such measures.

The system may have any one, or any combination insofar as they are compatible, of the following optional features.

The determination means may combine a number of calculated volumes, and use them in a neural network to provide an indication that the patient has a neurocognitive disease.

The parts of the limbic lobe for which the calculation means calculates a volume may include any one or more of, and preferably all of: the anterior cingulate gyrus, hippocampal and parahippocampal gyri, insula, and the superior temporal gyrus

The parts of the sub-cortical grey nuclei for which the calculation means calculates a volume may include any one or more of, and preferably all of: the amygdala, the caudate nucleus, the pallidum, and the thalamus.

The data obtaining means may be further configured to obtain data indicative of a physical structure of one or more parts of one or more, and preferably all of, of the following territories of the patient’s brain: central region, and temporal lobe. The part of the central region for which the calculation means calculates a volume may include the precentral gyrus. The parts of the frontal lobe for which the calculation means calculates a volume may include any one or more of, and preferably all of: the orbital surfaces of the frontal cortex and the medial surface areas of the frontal cortex. The part of the temporal lobe for which the calculation means calculates a volume may include the middle temporal gyrus.

The neurocognitive disease may be one of: Alzheimer’s disease and behavioural variant fronto-temporal dementia. Preferably the system and method can classify a patient as either: being cognitively normal; having Alzheimer’s disease, or having behavioural variant frontotemporal dementia. The system and method may use two binary determinations in classifying the patient, a first in which the patient is classified as either being cognitively normal or having Alzheimer’s disease, and a second in which the patient is classified as either having Alzheimer’s disease or having behavioural variant fronto-temporal dementia. Determining the indication that the patient has a neurocognitive disease may include determining whether the patient is likely to have behavioural variant fronto-temporal dementia, and if this is determined to be true, the method may perform a subsequent step of determining whether the patient has behavioural variant fronto-temporal dementia or Alzheimer’s disease. The determination whether the patient is likely to have bvFTD may comprise a step of classifying, preferably using a binary classifier, the comparisons of volumes as either being indicative of a bvFTD group or being indicative of a normal control or healthy group. The subsequent step may comprise a step of classifying, preferably using a binary classifier, the comparison of volumes as being either indicative of a bvFTD group or being indicative of an AD group. Alternatively, the indication that the patient has a neurocognitive disease may include a step of classifying, preferably using a trinary classifier, the comparison of volumes as being indicative of one of: a bvFTD; a normal control or healthy group; or an AD group.

The structural neurological data may be magnetic resonance imaging data.

The determination means may be a trained perceptron neural network. The perceptron neural network may be a single-layered perceptron network. The trained artificial neural network may take as an input a volume of the or each part of each territory, and provide as an output a classification indicative of whether the patient has or does not have a neurocognitive disease. The trained artificial neural network may take as an input one or more regional volumes of the or each part of each territory.

The calculation means may calculate the volume of the or each part of each territory using voxel-based-morphometry. The calculation of the volume may be performed either voxel- wise or based on an assumed parcellation of the brain. By voxel-wise, it may be meant that the calculation of volume is performed without an assumption as to the labelling of particular brain volumes, for example by the pair-wise comparisons of voxels of a control, reference group’s brain to the voxels of the patient’s group brain. A calculation of a degree of atrophy may be performed assuming specific labelling of particular brain anatomy (for example, after positive identification of the hippocampus, given the brain atlas, by the comparison of reference volumes for the hippocampus to the hippocampus volumes of the patients, at the group level). The calculation means may compare the volume of the or each part of each territory, voxel by voxel or by aggregation of voxels (labelled regions), to an expected value of a normal control sample or reference volume (the reference volume being calculated from a healthy or control group) to determine the degree of atrophy. The calculation means may use an output of the FreeSurfer software package running on one or more processors or central processing units.

Further aspects of the present invention provide: a computer program comprising code which, when run on a computer, causes the computer to perform the method of the first aspect; a computer readable medium storing a computer program comprising code which, when run on a computer, causes the computer to perform the method of the first aspect; and a computer system programmed to perform the method of the first aspect.

Brief Description of the Drawings

Embodiments of the invention will now be described by way of example with reference to the accompanying drawings in which:

Figure 1 shows a dendrogram created by the hierarchical agglomerative clustering, two hundred and thirteen bvFTD subjects are progressively linked together using Ward’s method with vertical lines indicating cut off at either four (blue line) or three (green line) clusters for which voxel-wise comparisons of MR images between the bvFTD and the healthy control group were performed;

Figure 2A shows regions of atrophy in bvFTD individuals (bvFTD volumes < expected NC volumes) clustered based on differences in the 68 regional volumes sagittal and medial views, yellow areas represent significant volume loss in each bvFTD cluster/sub-group based on pair-wise comparisons with the healthy control group, when corrected for total intracranial volume; Figure 2B shows 3D surface renderings showing patterns of grey matter atrophy of bvFTD clusters in comparison with the healthy controls when corrected for total brain volume;

Figure 3 shows pair-wise differences between the four identified bvFTD cl usters/g roups mapped onto the cortical surface, where hot/cold colours indicate t-test statistics used for the comparisons, and hot colours indicate ’more’ atrophy (as indicated in each panel by an inequality sign);

Figure 4 shows a plurality of schematic brain views (sagittal and middle) of the common-to- all (grey) and distinct (magenta) regional atrophy in the four bvFTD sub-groups, where atrophied regions were labelled using Automated Anatomical Labelling (AAL);

Figure 5 shows a plurality of box-plots with individual data points superimposed for behavioural and functional sub-scores in bvFTD sub-groups;

Figure 6 shows a 3D surface rendering showing neural correlates of cognitive sub-scores in the group of bvFTD patients, indicating that it is possible to map function and behavioural deficits to specific cortical regions;

Figure 7A shows a 3D scatter plot showing the separation of the four bvFTD subtypes using the first three components (PCA scores) in the factor reduction analysis on the 68 regions of interest used in hierarchical agglomerative clustering, the first three components explained 35 % of the total variance in the grey matter differences between the subjects, each of the subtypes occupied a specific area in the reduced factors’ space;

Figure 7B shows a 3D scatter plot showing separation of the four bvFTD sub-types using the first three components (PCA scores) in the factor reduction analysis on the cognitive and behavioural sub-scores, the first three components explain 55.4 % of the total variance in cognitive and behavioural outcomes between the subjects, however there is no clear separation between anatomical sub-types in the reduced space;

Figure 8 shows a plurality of box-plots with individual data points superimposed for cognitive sub-scores in bvFTD sub-groups;

Figure 9 shows a 3D surface rendering showing neural correlates of cognitive sub-scores in the group of bvFTD patients;

Figure 10 shows a method according to an embodiment of the present invention; and Figure 11 shows a system according to an embodiment of the present invention.

Detailed Description and Further Optional Features

Aspects and embodiments of the present invention will now be discussed with reference to the accompanying figures. Further aspects and embodiments will be apparent to those skilled in the art. All documents mentioned in this text are incorporated herein by reference

Methods

Study Participants

Study TRx-237-007 (ClinicalTrials.gov identifier: NCT01626378) was designed as a 52-week Phase 3, randomised, controlled, double-blind, parallel-group trial conducted in over 70 sites in Canada, the United States, Australia, Asia, and Europe. Eligible patients were those younger than 80 years of age, and diagnosed with bvFTD according to criteria revised by the International bvFTD Criteria Consortium (Rascovsky et al., 2011), with a Mini-mental State Examination (MMSE; Folstein et al., 1975) score greater than or equal to 20 at screening. An additional requirement was that patients had to meet the criterion of having definite brain atrophy (assessed visually rather than quantitatively) in frontal and/or temporal lobes scoring 2 or more on a scale proposed by Kipps et al. (Kipps et al., 2008). Concomitant use of acetylcholinesterase inhibitors or memantine (or both) was permitted provided this was at a stable dose for at least 18 weeks before randomisation. Concomitant use of serotonergic antidepressant, antipsychotic (except clozapine or olanzapine) and sedative medications was also permitted at stable doses where clinically feasible. Each patient had one or more study partners participate with them in the trial as informants. Patients were excluded from the study if they had a significant CNS disorder other than bvFTD. A detailed list of inclusion and exclusion criteria in the protocol is available in the Supplementary Materials in Shiells et al. (Shiells et al., 2020). Baseline MRI scans were evaluated by an independent neuroradiologist to determine eligibility.

In addition, MRI scans were obtained from 244 age-matched healthy controls from the Aberdeen 1936 Birth Cohort (ABC36) brain imaging database held in the Aberdeen Biomedical Imaging Centre at the University of Aberdeen.

MRI data collection and analysis The acquisition protocol was standardised across sites, and all data were centrally collected, quality-controlled, and analysed. MRI data acquisition included a 3D sagittal T1-weighted sequence which we use in our analysis here. Brain images were acquired using a 3D MPRAGE sequence or the specific manufacturer equivalent 3DT1 sequence. FreeSurfer version 5.3.0 (http://freesurfer.net) was used to extract regional volumes for subsequent clustering analysis. FreeSurfer automated segmentation parcellates the brain into 76 regions according to the Desikan-Killiany Atlas (Desikan et al., 2006). For the purpose of the study, 68 regional volumes (34 from each hemisphere) of the frontal, temporal or parietal and additional sub-lobar regions (limbic lobes, basal ganglia, amygdala, and thalamus) were selected. The full list of regions is as follows:

Amydala Parahippocampal

Caudal anterior cingulate Pars opercularis

Caudal middle frontal Pars orbitalis

Caudate Pars triangularis

Entorhinal Pericalcarine

Frontal pole Post central

Fusiform Posterior cingulate

Hippocampus Precentral

Inferior parietal Putamen

Inferior temporal Rostral anterior cingulate

Insula Rostral middle frontal

Isthmus of cingulate Superior frontal

Lateral orbitofrontal Superior parietal

Medial orbitofrontal Superior temporal

Middle temporal Temporal pole

Pallidum Thalamus

Paracentral Transversal temporal

Hierarchical agglomerative clustering, implemented in SPSS v.23.0, was used to classify differences/similarities in the 68 regional volumes. The bottom-up hierarchical agglomerative clustering is based on similarities and linkages between data points (subject-wise region of interest, ROI, volumes on MRI), with successive agglomeration of pairs of clusters until all clusters are merged into a single cluster containing all subjects. Similarity was measured by Euclidian distance between pairs of data points (each data point representing a brain volume of a specific region) and linkage was measured by Ward’s linkage method (Ward 1963).

Voxel-based morphometry (VBM) was used in parallel with ROI-based approaches. The VBM processing procedure employed followed the steps described in Ashburner (Ashburner 2015). In short, the images were first segmented into grey matter, white matter, and cerebrospinal fluid mask images (Ashburner and Friston 2005; Ashburner 2015). Each class of segmented images were then warped together and non-linearly registered so that they matched each other (Ashburner and Friston 2011). Images were normalised to the Montreal Neurological Institute space and smoothed with a Gaussian kernel (8 mm FWHM). Each group identified by the clustering technique was compared to the healthy control group. Regions showing atrophy in the bvFTD group were identified (by comparison with the NC sample) from the MNI coordinates of the voxels within the areas that were significantly different using maximum difference t-test statistics. All image processing steps and statistical analysis were implemented in the Statistical Parametric Mapping (SPM12) software package (available at http://www.fiLion.ucl.ac.uk/spm/). The t-tests were performed on each pair of voxels I volumes corrected for age, gender, and either estimated total intracranial volume (Whitwell et al., 2009) or total brain volume (Bigler and Tate 2001) to correct for global atrophy I severity. To correct for the false discoveries of significant differences due to multiple tests, the t-tests statistics were corrected at the significance level p < 0.05 using the family-wise-error correction available in the VBM statistical package.

Clinical assessments

Baseline clinical assessments included the Addenbrooke’s Cognitive Examination Revised (ACE-R) (Mioshi et al., 2006), the Mini-Mental State Examination (Folstein et al., 1975), the Frontotemporal Dementia Rating Scale (FRS) (Mioshi et al., 2010; Whitwell et al., 2009) and the Functional Assessment Questionnaire (FAQ) (Pfeffer et al., 1982). The bvFTD subtypes were compared using these scales and using subscales derived from the ACE-R and FRS prior to identification of the bvFTD subtypes. A total of 18 subscales were created: 11 cognitive subscales (from ACE-R) and 7 behavioural subscales (from the FRS). The FAQ score was used in its entirety as an independent measure of activities of daily living.

Statistical analysis Statistical analyses were performed using the SPSS v.23.0, employing paired samples t- tests to compare males and females in Table 1 ,

Significant differences between males and females: *p<0.05, **p<0.01. R = Right; L = Left; A= Ambidexterous; ADL = Activies of Daily Living

One-way ANOVA was used to test differences between bvFTD subtypes in cognitive and behavioural sub-scores as shown in Table 4 below.

Results

Demographic and clinical features of the populations studied

A total of 213 bvFTD patients of 220 randomised to the trail were included in the study, based on baseline MRI scan quality and complete clinical data required for the study at baseline. Baseline demographic and clinical data are provided in Table 1 above. Mean (± SD) age was 63 ± 7 years for both males (136) and females (77). Total years in education was 15.4 ± 0.5, with no difference between males and females. The estimated total intracranial volume (eTIV) was significantly larger in males (1600 ± 220 cm ³ ) than inn females (1400 ± 150 cm ³ ), although there was no difference in brain fraction (BF, 0.67 ± 0.06). The MMSE score was significantly higher in males (25.4 ± 3.5) than in females (22.9 ± 4.0), as was the total ACE-R score (males: 72 ± 16; females 62 ± 14). Males performed better on most of the ACE-R subscales apart from phonemics, language structure, episodic memory, and perceptual abilities. In contrast, there was no overall differences on either the FRS or the FAQ score between males and females. The only FRS subscales showing a gender difference were ADL (where males performed better) and disinhibition (where males performed worse). There were no demographic differences between patients prescribed symptomatic treatments approved for AD (but not bvFTD) and those not receiving these treatments. There were 133 males and 111 females in the healthy elderly group. There were no sex differences in age, years of education, or MMSE score. The healthy elderly group was significantly older (69 ± 2 years), had less education (11 ± 2 years), and had a higher MMSE score (28.9 ± 1 .2) than the bvFTD group.

Classification of bvFTD subjects by agglomerative clustering based on regional brain volumes

A hierarchical agglomerative clustering algorithm was applied using Euclidean distance and Ward linkage to provide measures of differences/similarities in the 68 regional volumes of the Desikan-Killiany Atlas (Desikan et al., 2006). The tree/dendrogram is shown in Figure 1. It is possible to classify bvFTD groups into either 3 or 4 clusters depending on the cluster distance. The 4-group clustering was used for further analysis. Each of these groups was then treated as a single group and compared to the healthy elderly group using VBM. Figure 2A shows the 3D surface rending of the voxel-wise differences between the bvFTD groups and the healthy elderly group after correction for total intra-cranial volume. A similar result was found when the correction was based on estimated brain volume (Figure 2B).

Four anatomical sub-types were designated based on this analysis and named according to the cortical territories showing a high degree of atrophy: fronto/temporo/parietal (FTP), frontal-dominant (FD), temporal-dominant (TD), and sub-lobar (SL). The differences in cortical atrophy across bvFTD sub-types are shown in Figure 3.

Degeneration of central basal and limbic nodes as a core feature of bvFTD

Having confirmed the classification of the bvFTD sub-types based on distinct patterns of cortical atrophy, it was then determined how these are linked to atrophy in the subcortical and limbic regions (Figure 4 and Table 3A, below). As shown in Tables 3A and 3B, it was found that it is possible to distinguish regions that are common to the four subtypes form those that are not. Table 3A illustrates the cortical regions with differences in the brain volume which are common to all four identified bvFTD sub-types, the regions are all bilateral. The group-wise differences at the whole brain level were classified by VBM against the healthy group using maximum of the t-test statistics (separated by more than 1 mm) within a cluster and then labelled according to the Automated Anatomical Labelling (AAL) function of SPM.

Regions /Sub-type-No Subjects (N) FTP (82) FD (41) SL (39) TD (51)

Table 3B illustrates the cortical regions with differences in the brain matter volume specific to either of the four identified bvFTD sub-types (black-filled circles). The group-wise differences at the whole brain level were identified as per Table 3A.

The limbic structures found to be common to all four sub-types in terms of atrophy include the anterior cingulate gyrus, hippocampal, and parahippocampal gyri, insula, and temporal pole (superior temporal gyrus). The subcortical grey nuclei affected in all subtypes included amygdala, caudate nucleus, pallidum, and thalamus. The cortical regions common to the subtypes were the orbital surface areas of the frontal cortex (inferior frontal gyrus, olfactory cortex, and gyrus rectus) and the medial surfaces of the frontal cortex (superior frontal gyrus and supplementary motor area). Only the middle temporal gyrus of the temporal lobe was shared between the sub-types. In all cases, the involvement of the common regions was bilateral.

A striking feature of the regions of atrophy shared by all sub-types is that they have all been postulated previously to be either members of the rich club (van den Heuval et al., 2012), functional binding nodes (Deco et al., 2017), or as regions having higher than average connectivity. As shown in Figure 4 and Table 3A, the rich club members identified as undergoing atrophy in all four bvFTD sub-types were superior frontal gyrus, thalamus, pallidum, putamen, and hippocampus. The functional ‘binding’ regions common to the four sub-types include the anterior cingulate and insula. In addition, the parahippocampal gyrus, amygdala, and caudate have been identified as highly connected nodes.

In contrast with these sub-type independent regions, the regions listed in Table 3B had more limited sub-type overlap and were generally unilateral. Atrophy in the superior temporal gyrus was unique to the TD sub-type. The FTP sub-type showed atrophy in the middle occipital gyrus and precuneus, and the latter is also seen in the FT sub-type. There was no overlap seen between atrophy and any of the rich club or linked regions that was unique to the SL sub-type. Denegation in the superior occipital lobe was unique to the frontal-dominant sub-type. Atrophy in the superior temporal gyrus, although a functional binding node, was unique to the temporal-dominant sub-type.

Cognitive, functional, and behavioural performance across bvFTD sub-types

The principle cognitive scales, ACE-R and MMSE, showed no significant differences according to the bvFTD sub-types as shown in Table 4 below. In contrast, the functional and behavioural scales, FAQ and FRS, showed significant differences (Table 4 and Figure 5), with the FD sub-type showing greater overall impairment than the others. To examine this further, FRS subscales were used to determine whether behavioural elements of the bvFTD syndrome could be differentiated according to subtype. As shown in Table 4 and Figure 5, a picture like that seen with the full scales emerged, namely that the FD sub-type was generally more impaired than the others. This could be seen for Behavioural Symptoms (characterised predominantly by lack of appropriate behaviours), Apathy/Disinterest and Disinhibition. The only significant exception was that both FD and TD sub-types were characterised by greater impairment on the subscales measuring Problematic Behaviours than the FTL and SL sub-types. Although it was possible to map function and behavioural deficits to specific cortical regions (Figure 6), these features did not discriminate between the bvFTD sub-types identified by structural criteria, apart from greater general impairment largely restricted to the FD sub-type (see Figure 7A and 7B).

Since overall cognitive impairment did not provide a basis for discriminating between the sub-types, it was then tested whether cognitive subdomains possessed greater discriminatory capacity. Here a more complex picture emerged, as shown in Table 4, Figure 8 and Figure 9. As expected, the TD sub-type was characterised more specifically be greater impairment in semantic memory and language specifics, but not episodic memory. The FD sub-type was differentiated by more prominent deficits in letter fluency. The FTP sub-type showed somewhat greater impairment in language phonemics and perceptual abilities. There were no cognitive deficits that could be linked more specifically to the SL sub-type. Although the FTD sub-type might appear to be more AD-like, there was no differences in likelihood of being prescribed symptomatic treatments approved for AD.

Clinical Characteristics FTP (N=82) FD (N=41) SL (N=39) TD (N=51) p -value

AChEI/Mem 64/18 34/7 29/10 44/7 0.48

Demographics

Age 63.5 (7.5) 63.3(7.4) 63.4 (7.3) 62.6 (7.8) 0.33

Education 14.6 (6.0) 16.0 (6.0) 15.7 (6.6) 15.0 (6.0) 0.30

Gender (M/F) 53/29 21/20 25/14 37/14

Table 4

The relationship between core features and cognitive, behavioural, and functional performance in bvFTD It is clear from the previous analysis that atrophy patterns in bvFTD can be split into those atrophy patterns which are common to all sub-types and those associated with a sub-type (Tables 3A and 3B). Using the core regions, a data reduction technique was performed and the first unrotated factor extracted which explained 40% of the variance found across all core regions. There was a significant correlation between this extracted factor and cognitive, behaviour and functional scores, see Table 5:

Next, a generalised linear model (GLM) approach was used to determine whether these associations are dependent on unique features of the sub-type. In this analysis, the cognitive, behaviour, and functional scores were considered separately as the dependent variable, and sex and sub-type were included as fixed factors, and age, the summary core factor and head size were included as covariates, see Table 6:

A post hoc analysis of the marginal mean of the sub-groups was also performed. The GLM analysis showed that the summary core feature drives the association with cognitive impairment and function (Activities of Daily Living, FAQ). By contrast, the behavioural scores are associated with the cortical sub-type, but not with the summary core features.

Discussion

A large clinical cohort of 213 well characterised bvFTD patients from a global, multicentre study have been analysed. The inventors aimed to examine how common and heterogeneous patterns of volumes account for clinical diversity in the syndrome. Four anatomically distinct sub-types were confirmed at the cortical level, and atrophy in regions of the sub-lobar and limbic sub-types have been identified as a core feature of bvFTD. A summary metric based on core region atrophy was found to associate significantly with cognitive and functional performance across all sub-types. Conversely, cortical heterogeneity was associated with behavioural performance independent of the variance explained by the core features. Therefore, bvFTD can now be understood as comprising a core disturbance in the highly connected subcortical and limbic brain structures that is closely linked to cognitive and functional impairment.

Despite the existence of the four anatomically distinct patterns of cortical atrophy in the population which have been analysed, there is a homogenous pattern of atrophy across subcortical and limbic regions that is common to the anatomical sub-types. Of 39 brain regions showing atrophy when compared to healthy subjects, 16 regions were found to be common to all four sub-types, whereas 23 had selective sub-type associations. Of the 16 sub-type independent regions of atrophy, 10 regions are known to be brain hubs, i.e., brain regions with higher than average connectivity. These 10 regions map either to the so called ‘rich- club’ of highly connected nodes, to highly connected functional binding nodes, or nodes known to have higher than average connectivity in either functional or structural MRI studies. These network hubs are central to communication and functional integration of the brain and represent potential hotspots for impaired connectivity across multiple brain networks. It is known from previous work that the sub-networks of highly connected nodes play an important role in efficient information processing between segregated brain areas (Bullmore and Sporns 2012) and have been found to be associated with cognitive performance (Baggio et al., 2015) in healthy brains. Meta-analysis of MRI studies has suggested that there is a high vulnerability of the structural brain hubs and their connections in AD (Crossley et a/., 2014; Sha et al., 2017) although the hubs implicated in bvFTD and AD differ (Daianu et al., 2016). Of the 23 regions where atrophy was not shared across the sub-types, two were found to be either subcortical or limbic, and five are members of the rich club or functional binding group. It can be concluded therefore that degeneration in basal, limbic, and frontal networks that have high levels of connectivity represents a core feature of the bvFTD syndrome irrespective of the anatomically distinct sub-types described above.

Degeneration of brain network hubs is not unique to bvFTD. As noted above, the anatomical sites of atrophy differ between different neurodegenerative disorders. For example, the central brain regions affected in AD are more likely to be the medial temporal and parietal regions, although the thalamus and hippocampus are consistently atrophied in both bvFTD and AD. It has been proposed that the increased traffic that hubs are required to support may help to explain why these regions have preferentially greater vulnerability or neurological disorders more generally (Stam 2014; van den Heuvel and Sporns 2013). A high degree of connectivity may also make certain regions more vulnerable to prion-like spread of pathology arising stochastically in linked sub-regions. These need not be mutually exclusive, since a chronically high level of activity may itself lead to high demands on turnover of vulnerable protein systems and predisposed to pathological aggregation and transmission. The results shown here support the existence of a common underlying pattern of degeneration, which is not restricted to the salience network (as considered previously to be the case). Different sub-types of bvFTD, which can be distinguished at the cortical level, ranging from absence of cortical lobar atrophy, to lobe-specific dominance, to multi-lobar atrophy, all share degeneration in the basal, limbic, and frontal networks described herein.

In the cohort under investigation, the fronto-temporo-parietal sub-type had the highest frequency (39%) and the frontal-dominant (19%), the temporal-dominant (24%), and the sub- lobar (18%) sub-types had comparable lower frequencies. Therefore, the syndrome as defined by consensus clinical criteria and by the requirement for a significant degree of frontal and/or temporal lobe atrophy remains neuroanatomically heterogeneous in the population studied. The results corroborate two smaller studies reporting differences in patterns of degeneration across cortical areas in patients diagnosed as having bvFTD by consensus criteria (Ranasinghe et al., 2016; Whitwell et al., 2009). This agreement across studies is preserved despite the sampling of different subsets of regional volumes and utilisation of different statistical classifications. Whilst the sub-types identified herein find partial matches in Ranasinghe and Whitwell, atrophy in the specific regions identified with respect to Table 3A have not been identified in the literature as a key diagnostic criterion for certain neurological disorders. Ranasinghe postulated that the sub-lobar sub-type represents a true bvFTD sub-type which progresses more slowly. The analysis performed herein shows no global, cognitive, functional, or behaviour differences which might have been expected if it did represent an earlier stage of the disease. The data presented herein support the concept that the sub-lobar sub-type is indeed a distinct sub-type, and its prevalence is comparable to that of the frontal- and temporal-dominant subtypes.

Heterogeneity at the cortical level is associated to only a limited extent with distinct behavioural, functional and cognitive features. The frontal-dominant sub-type is characterised by greater global impairment on both the FRS and FAQ scales. This is surprising, given that functional deficits are also commonly seen in AD where the pattern of atrophy is predominantly temporo-parietal. It is also surprising that the frontal-dominant subtype is the most severely affected in terms of behavioural symptoms such as lack of appropriate social response, apathy, and disinterest, as well as disinhibition and problematic positive behaviours. In other words, in contrast to the overall importance of highly connected networks in defining the bvFTD syndrome, the frontal lobe remains particularly important for regulation of behaviour. In contrast, cognitive deficits segregate as expected, with the temporal-dominant form associated particularly with semantic memory and language semantics, and a stronger association between frontal-dominant atrophy and impairment in letter fluency.

Independently of the sub-type, a summary metric variable for the core features (the first unrotated factor), was found to have a highly significant statistically association with cognitive impairment, particularly with ACE-R, and with functional impairment as measured by the FAQ scale. After adjusting for the core factor, there was no residual association with the sub-type. On the other hand, the behavioural subscale derived from the FRS retained a significant association with anatomical sub-type after taking account for the core factor variable: the frontal sub-type showing statistically significant differences with respect to the temporal and sub-lobar sub-types. This goes some way to explaining the possible latent nature of the pure sub-lobar sub-type in which prominent behavioural deficits may not be demonstrated.

This mapping of imaging features to clinical features should be understood in the context of the inclusion criteria of the study, namely the requirement for present of brain in atrophy in frontal and/or temporal lobes scoring 2 or more on the Kipps scale. This means subjects with little or no frontal and/or temporal atrophy but who fulfilled all other criteria for bvFTD were not included. The results of the analysis discussed herein show that the diagnostic utility of MRI in the differentiation of bvFTD and/or AD from healthy controls and dementias may be best served by examining the core features with or without the frontal and temporal lobes.

Application for diagnosis

Following the analysis discussed above, a neural network was trained using the core features identified which is capable of discriminating bvFTD from controls and AD defined patients.

Method - Data Sources

Three data sources were used in the creation of the diagnostic tool:

[1] bvFTD data obtained from the TRx-237-007 study discussed above;

[2] AD data obtained from TauRx Study TRx-237-005 which was a 18 month phase 3 randomized, controlled, double-blind, parallel-group study at 115 academic centres and private research clinics across 16 countries in Europe, North America, Asia, and Russia. Eligible patients were those younger than 90 years of age with a diagnosis of mild to moderate probably Alzheimer’s disease according to criteria from the National Institute of Aging and the Alzheimer’s Association, an MMSE score of 14 - 16 inclusive, and a Clinical Dementia Rating (CDR) total score of 1 or 2. Baseline MRI scans were evaluated by an independent neuroradiologist to determine eligibility.

[3] Normal control (NC) data obtained from healthy elderly control cohort Aberdeen 1936 Birth cohort. Method - Image Acquisition

MRI data acquisition from each source included a 3D sagittal T1 -weighted sequence which was used in the analysis. A full description of the acquisition protocol can be found in the MRI protocol for the ADNI project (http://adni.loni.usc.edu/). Brain images were acquired using a 3D MPRAGE (or equivalent) sequence of the specific manufacture equivalent 3DT1 sequence.

Method - Al Approach

The automatic architecture section approach, available within the SPSS software package version 25 was used. The perception architecture consisted of 1 layer with a variable number of nodes (for example, 6 nodes) and a bias node depending on the task. A scale conjugate gradient approach was used. Unless otherwise stated, the SPSS default values were used.

Method - Al Discovery and Validation

In an initial step, the ABC36 [3] and Tau bvFTD [1] samples were split into two equal subgroups randomly. These were labelled the discovery set (NC=100, bvFTD=121) and the validation set (N=113, bvFTD=124) for bvFTD or NC classification.

In a second step, the AD [2] and Tau bvFTD [1] samples were split into two equal subgroups randomly. These were also labelled the discovery set (bvFTD=93, AD=347) and the validation set (bvFTD = 120, AD=336) for bvFTD or AD classification.

Method - Image Processing

Volumetric region of interest values for each imaging dataset were extracted using FreeSurfer.

Preliminary group comparisons

In a similar manner to that described above, a comparison was made for bvFTD [1] subtypes with NC [3] for each ROI volume. This comparison was repeated replacing the NC with AD [2] subjects. The object of these comparisons was to establish the core atrophic features of bvFTD when compared to NC and AD patients. Table 7 shows the regions where all subtypes are significantly different (p < 0.05) from either NC or AD or both.

Table 7 bvFTD or NC classification

A first Al task was classification of bvFTD or NC using the NC significant regions of interest (Table 7). 337 samples were provided in the training set, and 123 in the testing set for a total of 458 samples.

The classifier demonstrated an area under the ROC curve of 0.998. A confusion matrix for the classifier performance on the training and testing data sets is shown below:

Classification

Sample Observed Predicted

Dependent Variable: bvFTDorNot

Therefore, in the testing set, the classifier achieved an 95% true positive score for bvFTD classification, and a 91 .8% true positive score for NC classification. A second Al task was classification of bvFTD or AD using the AD significant regions of interest Table 7. 302 samples were provided in the training set, and 141 in the testing set for a total of 443 samples.

The classifier demonstrated an area under the ROC curve of 0.957. A confusion matrix for the classifier performance on the training and testing data sets is shown below:

Classification

Predicted

Therefore, in the testing set, the classifier achieved a 91 .7% true positive score for AD classification, and a 85.5% true positive score for bvFTD classification. A third Al task was classification of bvFTD or Not (AD or NC using) the ROIs significant to both AD and NC (see Table 7) and not the regions significant to AD or NC. 315 samples were provided in the training set, and 141 in the testing set, for a total of 456 samples. The model was a single layer with 2 nodes.

The classifier demonstrated one versus all area under the ROC curve values of 0.975.

A confusion matrix for the classifier performance on the training and testing data sets is shown below:

Classification

Predicted

Dependent Variable: bvFTDorNot

Therefore, in the testing set, the classifier achieved a 91 .0% true positive score for Not bvFTD classification, a 93.2% true positive score for bvFTD.

An example Al task is presented below (corresponding to the last Al task above), as are the corresponding weights estimated: /STOPPINGRULES ERRORSTEPS= 1 (DATA=AUTO) TRAININGTIMER=ON (MAXTIME=15) MAXEPOCHS=AUTO ERRORCHANGE=1. OE-4 ERRORRATIO=0.001 /MISSING USERMISSING=EXCLUDE .

Parameter Estimates

Accordingly, it has been shown that an appropriately trained machine classifier can distinguish between bvFTD, AD, and normal cognitive based on a feature analysis including the specific territories of the brain identified above.

Figure 10 shows a method according to the present invention. In a first step, S1 , structural volumetric neurological data are obtained. The structural neurological data is indicative of a physical structure of one or more parts of at least the following territories of a patient’s brain: limbic lobe and sub-cortical grey nuclei. In some examples the neurological data is indicative of the physical structure of all of the parts listed in Table 3A above. Next, in step S2, the volume xyis calculated for the i ^th part of the j ^th territory. For example, the hippocampal gyrus of the limbic lobe. The method then moves to step S3, where it is determined if the volume of all parts of all regions or territories have been calculated. If not, the method returns to steps S2 and the volume is calculated for the next part.

Once the volume has been calculated for all parts of all regions, the method moves to step S4, where it is determined, using each volume, an indication as to whether the patient has a neurocognitive disease. In an alternative example, the calculated volume(s) are used to determine a degree of atrophy in the corresponding part(s) of each region or territory. In such an example, the calculated volume is compared to a expected reference value for the corresponding part, and a determination as to a degree of atrophy is made. This determined degree of atrophy can then be used as before to determine an indication as to whether the patient has a neurocognitive disease.

Figure 11 shows a system 100 according to the present invention. The system includes a network adaptor 101 , a first input/output adaptor 102, a second input/output adapter 103, a CPU 104, and memory 105. The memory contains processor executable instructions which, when performed on the CPU 104, cause the CPU to perform the method shown in Figure 10. The data may be obtained either from an external network 106, directly from an MRI machine 107, or from local storage 108. It will of course be appreciated that where the system 100 relies only on data from, for example, the external network 106 it does not contain the first or second input/output adapters. The CPU 104, once it has determined the indication as to whether the patient has a neurocognitive disease, may present this on a display (not shown) to transmit the indication to the external network 106. In this manner, clinics may transmit raw or partially processed structural neurological data to the system, whereupon the system will determine an indication as to whether the patient has a neurocognitive disease and transmits this indication back.

The systems and methods of the above embodiments may be implemented in a computer system (in particular in computer hardware or in computer software) in addition to the structural components and user interactions described.

The term “computer system” includes the hardware, software and data storage devices for embodying a system or carrying out a method according to the above described embodiments. For example, a computer system may comprise a central processing unit (CPU), input means, output means and data storage. Preferably the computer system has a monitor to provide a visual output display. The data storage may comprise RAM, disk drives or other computer readable media. The computer system may include a plurality of computing devices connected by a network and able to communicate with each other over that network.

The methods of the above embodiments may be provided as computer programs or as computer program products or computer readable media carrying a computer program which is arranged, when run on a computer, to perform the method(s) described above.

The term “computer readable media” includes, without limitation, any non-transitory medium or media which can be read and accessed directly by a computer or computer system. The media can include, but are not limited to, magnetic storage media such as floppy discs, hard disc storage media and magnetic tape; optical storage media such as optical discs or CD- OMs; electrical storage media such as memory, including RAM, ROM and flash memory; and hybrids and combinations of the above such as magnetic/optical storage media.

While the invention has been described in conjunction with the exemplary embodiments described above, many equivalent modifications and variations will be apparent to those skilled in the art when given this disclosure. Accordingly, the exemplary embodiments of the invention set forth above are considered to be illustrative and not limiting. Various changes to the described embodiments may be made without departing from the spirit and scope of the invention.

In particular, although the methods of the above embodiments have been described as being implemented on the systems of the embodiments described, the methods and systems of the present invention need not be implemented in conjunction with each other, but can be implemented on alternative systems or using alternative methods respectively.

References

Ashburner, J. Vbm tutorial. University College London 2015; http://www.fil.ion.ucl.ac.uk/john/misc/VBMclass15.pdf

Ashburner, J, Friston, KJ. Unified segmentation. Neuroimage 2005; 26: 839-51

Ashburner, J, Friston, KJ. Diffeomorphic registration using geodesic shooting and Gauss- Newton optimisation. Neuroimage 2011 ; 55: 954-67 Baggio, HC, Segura, B, Junque, C, de Reus, MA, Sala-Llonch, R, Van den Heuvel, MP. Rich club organization and cognitive performance in healthy older participants. J Cognitive Neurosci 2015; 27: 1801-10

Bigler, ED, Tate, DF. Brain volume, intracranial volume, and dementia. Investigative Radiol 2001 ; 36: 539-46

Bullmore, E, Sporns, O. The economy of brain network organization. Nat Rev Neurosci 2012; 13: 336-49

Crossley, NA, Mechelli, A, Scott, J, Carletti, F, Fox, PT, McGuire, P, et al. The hubs of the human connectome are generally implicated in the anatomy of brain disorders. Brain 2014; 137: 2382-95

Daianu, M, Mezher, A, Mendez, MF, Jahanshad, N, Jimenez, EE, Thompson, PM. Disrupted rich club network in behavioral variant frontotemporal dementia and early-onset Alzheimer's disease. Human Brain Mapping 2016; 37: 868-83 de Haan, W, Pijnenburg, Y, Strijers, R, van der Made, Y, van der Flier, W, Scheltens, P, et al. Functional neural network analysis in frontotemporal dementia and Alzheimer's disease using EEG and graph theory. BMC Neurosci 2009; 10: 101

Deco, G, Van Hartevelt, TJ, Fernandes, HM, Stevner, A, Kringelbach, ML. The most relevant human brain regions for functional connectivity: Evidence for a dynamical workspace of binding nodes from whole-brain computational modelling. NeuroImage 2017; 146: 197-210

Desikan, RS, Segonne, F, Fischl, B, Quinn, BT, Dickerson, BC, Blacker, D, et al. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage 2006; 31 : 968-80.

Folstein, MF, Folstein, SE, McHugh, PR. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12: 189-98

Josephs, KA, Whitwell, JL, Knopman, DS, Boeve, BF, Vemuri, P, Senjem, ML, et al. Two distinct subtypes of right temporal variant frontotemporal dementia. Neurology 2009; 73: 1443-50 Kipps, CM, Nestor, PJ, Dawson, CE, Mitchell, J, Hodges, JR. Measuring progression in frontotemporal dementia: Implications for therapeutic interventions. Neurology 2008; 70: 2046-52.

Lijegren, M, Landqvist Waldo, M, Frizell Santillo, A, Ullen, S, Rydbeck, R, Miller, B, et al. Association of neuropathologically confirmed frontotemporal dementia and Alzheimer disease with criminal and socially inappropriate behavior in a Swedish cohort. JAMA Network Open 2019; 2: e190261

Mendez, MF. The unique predisposition to criminal violations in frontotemporal dementia. J Am Acad Psychiatry Law 2010; 38: 318-23

Mioshi, E, Dawson, K, Mitchell, J, Arnold, R, Hodges, JR. The Addenbrooke's Cognitive Examination Revised (ACE-R): a brief cognitive test battery for dementia screening. Int J Geriat Psychiatry 2006; 21 : 1078-85

Mioshi, E, Hsieh, S, Savage, S, Hornberger, M, Hodges, JR. Clinical staging and disease progression in frontotemporal dementia. Neurology 2010; 74: 1591-97

Pfeffer, Rl, Kurosaki, TT, Harrah, CH, Chance, JM, Filos, S. Measurement of functional activities in older adults in the community. J Gerontol 1982; 37: 323-29

Ranasinghe, KG, Rankin, KP, Pressman, PS, Perry, DC, Lobach, IV, Seeley, WW, et al. Distinct subtypes of behavioral variant frontotemporal dementia based on patterns of network degeneration. JAMA Neurol 2016; 73: 1078-88

Rascovsky, K, Hodges, JR, Kipps, CM, Johnson, JK, Seeley, WW, Mendez, MF, et al. Diagnostic criteria for the behavioral variant of frontotemporal dementia (bvFTD): Current limitations and future directions. Alzheimer Dis Assoc Disord 2007; 21 : S14-18

Rascovsky, K, Hodges, JR, Knopman, D, Mendez, MF, Kramer, JH, Neuhaus, J, et al. Sensitivity of revised diagnostic criteria for the behavioural variant of frontotemporal dementia. Brain 2011 ; 134: 2456-77.

Rohrer, JD. Behavioural variant frontotemporal dementia-defining genetic and pathological subtypes. J Mol Neurosci 2011 ; 45: 583-88 Rohrer, JD, Guerreiro, R, Vandrovcova, J, Uphill, J, Reiman, D, Beck, J, et al. The heritability and genetics of frontotemporal lobar degeneration. Neurology 2009; 73: 1451-56

Seeley, WVV, Menon, V, Schatzberg, AF, Keller, J, Glover, GH, Kenna, H, et al. Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 2007; 27: 2349-56.

Sfera, A, Osorio, C, Gradini, R, Price, A. Neurodegeneration behind bars: from molecules to jurisprudence. Front Psychiat 2014; 5: 115

Sha, Z, Xia, M, Lin, Q, Cao, M, Tang, Y, Xu, K, et al. Meta-connectomic analysis reveals commonly disrupted functional architectures in network modules and connectors across brain disorders. Cerebral Cortex 2017; 28: 4179-94

Shiells, HC, Schelter, BO, Baddeley, TC, Rubino, CM, Ganesan, H, Hammel, J, et al.

Reduction in clinical decline and brain atrophy are concentration-dependent for leuco- methylthioninium bis(hydromethanesulphonate) (LMTM) in a Phase 3 trial in behavioural variant frontotemporal dementia. J Alzheimers Dis 2020

Stam, CJ. Modern network science of neurological disorders. Nat Rev Neurosci 2014; 15: 683-95 van den Heuvel, MP, Sporns, O. An anatomical substrate for integration among functional networks in human cortex. J Neurosci 2013a; 33: 14489-500 van den Heuvel, MP, Kahn, RS, Goni, J, Sporns, O. High-cost, high-capacity backbone for global brain communication. Proc Natl Acad Sci USA 2012; 109: 11372-7

Ward, JH. Hierarchical grouping to optimize an objective function. J Am Stat Assoc 1963; 58: 236-44.

Whitwell, JL, Xu, J, Mandrekar, J, Boeve, BF, Knopman, DS, Parisi, JE, et al. Frontal asymmetry in behavioral variant frontotemporal dementia: clinicoimaging and pathogenetic correlates. Neurobiol Aging 2013; 34: 636-39 Whitwell, JL, Przybelski, SA, Weigand, SD, Ivnik, RJ, Vemuri, P, Gunter, JL, et al. Distinct anatomical subtypes of the behavioural variant of frontotemporal dementia: a cluster analysis study. Brain 2009; 132: 2932-46

Zhou, J, Greicius, MD, Gennatas, ED, Growdon, ME, Jang, JY, Rabinovici, GD, et al. Divergent network connectivity changes in behavioural variant frontotemporal dementia and Alzheimer’s disease. Brain 2010; 133: 1352-67.

All references referred to above are hereby incorporated by reference.