Structural correlates of cognitive impairment in normal pressure hydrocephalus

Objectives The pathological bases for the cognitive and neuropsychiatric symptoms in normal pressure hydrocephalus (NPH) have not been elucidated. However, the symptoms may indicate dysfunction of subcortical regions. Previously, volume reductions of subcortical deep grey matter (SDGM) structures have been observed in NPH patients. The present study used automated segmentation methods to investigate whether SDGM structure volumes are associated with cognitive and neuropsychiatric measures. Methods Fourteen NPH patients and eight healthy controls were included in the study. Patients completed neuropsychological tests of general cognition, verbal learning and memory, verbal fluency and measures of apathy and depression pre‐ and postshunt surgery. Additionally, patients underwent 3 Tesla T1‐weighted magnetic resonance imaging at baseline and 6 months postoperatively. Controls were scanned once. SDGM structure volumes were estimated using automated segmentation (FSL FIRST). Since displacement of the caudate nuclei occurred for some patients due to ventriculomegaly, patient caudate volumes were also estimated using manual tracing. Group differences in SDGM structure volumes were investigated, as well as associations between volumes and cognitive and neuropsychiatric measures in patients. Results Volumes of the caudate, thalamus, putamen, pallidum, hippocampus and nucleus accumbens (NAcc) were significantly reduced in the NPH patients compared to controls. In the NPH group, smaller caudate and NAcc volumes were associated with poorer performance on neuropsychological tests and increased severity of neuropsychiatric symptoms, while reduced volume of the pallidum was associated with better performance on the MMSE and reduced apathy. Conclusions Striatal volume loss appears to be associated with cognitive and neuropsychiatric changes in NPH.

once. SDGM structure volumes were estimated using automated segmentation (FSL FIRST). Since displacement of the caudate nuclei occurred for some patients due to ventriculomegaly, patient caudate volumes were also estimated using manual tracing.
Group differences in SDGM structure volumes were investigated, as well as associations between volumes and cognitive and neuropsychiatric measures in patients.

Results:
Volumes of the caudate, thalamus, putamen, pallidum, hippocampus and nucleus accumbens (NAcc) were significantly reduced in the NPH patients compared to controls. In the NPH group, smaller caudate and NAcc volumes were associated with poorer performance on neuropsychological tests and increased severity of neuropsychiatric symptoms, while reduced volume of the pallidum was associated with better performance on the MMSE and reduced apathy.

Conclusions:
Striatal volume loss appears to be associated with cognitive and neuropsychiatric changes in NPH.

| INTRODUC TI ON
Normal pressure hydrocephalus (NPH) is characterized by a build-up of cerebrospinal fluid (CSF) in the brain which causes ventriculomegaly despite apparently normal CSF pressure at lumbar puncture.
Symptoms include gait apraxia, urinary incontinence and cognitive decline. 1 In addition, recent research suggests a high prevalence of apathy in NPH patients (65-86%). [2][3][4] NPH is treated via surgical CSF diversion which results in clinical and neuropsychological improvement for many patients. 5 The pathological mechanisms underlying the cognitive and neuropsychiatric symptoms in NPH are not well established. The executive dysfunction in NPH may suggest impaired frontal lobe functioning, 6 however, there is inconsistent evidence for frontal involvement from imaging studies of regional cerebral blood flow (rCBF) in NPH. 6,7 The pattern of cognitive decline is consistent with dysfunction of subcortical structures and may indicate disruption to the subfrontal white matter, limbic connections or connections between the frontal lobe and the subcortical deep grey matter (SDGM) structures. 8 Indeed, impaired rCBF has been found in the periventricular white matter and in the basal ganglia. 7,9 The presence of apathy (impaired goal-directed behaviour) in NPH is consistent with fronto-subcortical pathology. Apathy can result from lesions affecting fronto-subcortical circuitry, 10 or from focal lesions of the basal ganglia, 10 including lesions located in the caudate nucleus. 11 Diminished caudate volume was previously observed in NPH patients using voxel-based morphometry and hypothesized to contribute to the clinical symptoms of cognitive impairment and apathy. 12 However, this association was not examined.
We previously found evidence for an association between greater bicaudate ratio (greater ventriculomegaly) with increased levels of apathy and depression in NPH patients. 13 Additionally, evidence for associations between apathy and rCBF in the anterior cingulate cortices and the right caudate nucleus has been reported. 14 Subcortical volumes in NPH have not been extensively investigated using quantitative MRI techniques. The periventricular regions may be distorted due to ventriculomegaly. Atrophy may also occur over time due to ischaemia and subsequent cell death. 15 The few studies that have examined SDGM volumes in NPH focused only on individual structures or did not relate volumetric data to clinical information. 12,16,17 In the present study, we conducted volumetric assessments of a range of SDGM structures and investigated whether volumes are related to cognitive or neuropsychiatric measures.
Subcortical volumes were estimated using automated segmentation techniques. However, in some of the patients there was displacement of the caudate nuclei due to the degree of ventriculomegaly.
For this reason, we also used manual tracing techniques to estimate the volume of the caudate in all patients to supplement the analyses based on automated segmentation of the caudate.

| Participants
Sixteen patients with NPH were recruited between October 2007 and June 2009. Patients gave informed consent prior to enrolling in the study. Patients were assessed and scanned before and a mean of 6.4 months following shunt surgery. Patients were referred to be included in the study by a neurosurgeon (JDP) based on the presence of a clinical picture of gait disturbance, slowing of mentation and/ or short-term memory impairment. 6 One patient was excluded from the following analyses as they had comorbid Alzheimer's disease.
One additional patient was excluded due to a complicated postoperative clinical course.
Healthy controls were recruited via verbal screening at the CSF multidisciplinary clinic for inclusion criteria, and via advertisement in the local area. Interested partners or spouses of study participants who met the inclusion criteria were also offered healthy volunteer scans. Some healthy volunteers self-referred after becoming aware of the NPH imaging programme via the patient community.
Exclusion criteria for the healthy control group were (i) a diagnosis of a neurological disorder, (ii) the presence of contraindications to MRI scanning and (iii) age <60. Eight healthy controls (HCs) were scanned at one time-point.

| Neuropsychological assessment
Patients completed a brief battery of neuropsychological tests, selected on the basis of their sensitivity to NPH, 5 pre-and 089589/Z/09/Z), the MRC/Wellcome Trust Behavioural and Clinical Neuroscience Institute (joint award G00001354), the Human Brain Project, the NIHR Brain Injury MedTech in vitro Diagnostic Co-operative, and the NIHR Cambridge Biomedical Research Centre (Mental Health and Neurodegeneration Themes).

K E Y W O R D S
apathy, cognition, neuroimaging, neuropsychology, normal pressure hydrocephalus postoperatively. Global cognitive functioning was assessed using the Mini-Mental State Examination (MMSE). 18 Verbal learning and memory was assessed using the Hopkins Verbal Learning Test (HVLT). 19 Phonemic and semantic fluency were assessed using the Controlled Oral Word Association Test. 20 Apathy was assessed using the Cambridge-developed State Apathy Evaluation scale (self-rated; AES-S), 21 which measures state-related changes in apathy. Depression was assessed using the short form of the Geriatric Depression Scale (GDS-15). 22 IQ was assessed using the National Adult Reading Test. 23 To minimize practice effects, alternate versions of phonemic and semantic fluency categories and the HVLT word list were used at follow-up.

| MR imaging protocol
MR imaging was performed on a 3 T Siemens Tim Trio using a 12channel RF receive coil. The MR imaging protocol included a PDT2, FLAIR, MPRAGE and DTI. The T1-weighted structural sequence (MPRAGE) was acquired with TR/TE of 2300/2.98 ms, with a resolution of 1x1x1mm.

| Automated segmentation
Before processing, all images were manually reviewed for artefact. SDGM volumes were estimated for both groups (including at preand postshunt for NPH patients as the shunt artefact was above the level of the SDGM structures). Subcortical volumes (caudate, nucleus accumbens (NAcc), pallidum, putamen and thalamus) as well as volumes of the hippocampus and the amygdala were extracted from 3D T1-weighted images using an automated subcortical registration and segmentation tool in FSL (FSL FIRST; www.fmrib.ox.ac. uk/fsl). Examples of subcortical segmentation are shown in Figure 1.
All automated segmentations were thoroughly individually reviewed for accuracy. SDGM volumes were normalized for head size by multiplying by the volumetric scaling factors obtained using SIENAX, 24 part of FSL. 25 SIENAX starts by extracting brain and skull images from the single whole-head input data. 26 The brain image is then affine registered to MNI152 space (using the skull image to determine the registration scaling) 27 ; primarily in order to obtain the volumetric scaling factor to be used as a normalization for head size. Next, tissue-type segmentation with partial volume estimation is carried out, 28 in order to calculate total volume of brain tissue. As laterality effects were not hypothesized, normalized right and left volumes of the SDGM structures were combined to yield a single total volume for each structure.

| Caudate tracing
Manual tracing was performed on T1-weighted images using ITK-SNAP Finally, a 3D view of the segmented caudate nucleus was compiled and checked for abnormalities ( Figure 2). Volumes were normalized for head size by multiplying by the volumetric scaling factors from SIENAX.
Normalized right and left caudate volumes were combined to yield a single total volume.

| Statistical analysis
Data were analysed using the Statistical Package for the Social Sciences
Baseline mean (SD) age was 76.4 (3.9), range = 70-84, IQ was 110.8 Individual patient details are shown in Table 1. Two patients had comorbid depression (patients 6 and 9 in Table 1). Four HCs were male.

| Automated segmentation
The following analyses are based on automated subcortical seg- volumetric data for each structure were available for a limited number of patients. Therefore, correlational analyses were conducted at two time-points (pre-and postshunt) rather than on preto postshunt change scores.
Automated segmentation was successful in the HCs and all remaining patients so that analyses of SDGM volumetric data is based on 11 patients preoperatively and 12 postoperatively (apart from data for caudate volume which is based on eight patients preoperatively and 10 postoperatively).

| Subcortical volumes
Mean volumes of the SDGM structures for both groups are shown in Table 2. Volumes of all SDGM structures except the amygdala were significantly smaller in the NPH group at both pre-and postshunt compared to controls. These remained significant after controlling for age and sex.

| Neuropsychological profile pre-and postshunt surgery
Neuropsychological test scores for the NPH patients are shown in Table 3. Performance on semantic fluency, HVLT immediate and HVLT learning was significantly improved following shunt surgery.
In addition, there was a trend for improvement in all remaining test scores, except for GDS-15.

| Correlation of volumetric data with neuropsychology scores
The significance value for the Shapiro-Wilk test was less than 0.05 for the following: pre-shunt MMSE score, pre-shunt semantic fluency score, pre-shunt HVLT delayed score, postshunt MMSE score, postshunt AES-S and postshunt HVLT delayed score, suggesting that these data are not normally distributed. As such, Spearman's correlation coefficient (r s ) was used to investigate associations with these variables.

Pre-shunt
NAcc volume was significantly positively correlated with HVLT immediate (r = 0.70, P = 0.035). That is, greater volume was associated with better performance. There were no other significant correlations between SDGM volumes and cognitive test scores at pre-shunt. That is, greater pallidum volume was associated with poorer performance on the MMSE and increased apathy.    . This difference was statistically significant for the postshunt volumes only, t (9) = 6.90, P < 0.001. Patients were split according to whether they were included in or excluded from (due to unsuccessful caudate segmentation) the automated segmentation analyses. The mean manually segmented caudate volumes did not differ significantly between the two groups at pre-(P > 0.05) or postshunt (P > 0.05). Therefore, the smaller volumes obtained via manual tracing are not due to the inclusion of additional participants.

| Correlation of volumetric data with neuropsychology scores
Pre-shunt

| D ISCUSS I ON
We conducted a volumetric assessment of subcortical structures in NPH patients using automated segmentation. Manual tracing was conducted to supplement the caudate volume data. Caudate volumes obtained via automated and manual segmentation were significantly correlated at both pre-and postshunt suggesting that automated segmentation was satisfactory for the majority of patients.
All of the SDGM structures apart from the amygdala were significantly reduced in the NPH patients compared to healthy controls, and remained so after adjusting for age and sex. We conducted correlations to investigate associations between SDGM structure volumes and neuropsychological test performance. Using the data derived from automated segmentation, only NAcc volume was significantly associated with HVLT immediate score at preshunt; while at postshunt, larger caudate and NAcc volumes were associated with better performance on the MMSE, and reduced self-rated apathy. Caudate volume was also positively associated with semantic fluency at postshunt; and NAcc with verbal learning and memory. In addition, greater volume of the pallidum at postshunt was associated with poorer performance on the MMSE and increased self-rated apathy. The association between pallidum volume and scores is in the unexpected direction and should be clarified in a larger sample.
Since the analyses relating to caudate volume were based on small (and differing) Ns manual caudate tracing was also conducted. In summary, while most of the SDGM structures investigated in the present study were significantly smaller in NPH patients compared to controls, striatal volume loss appears to be associated with the cognitive and neuropsychiatric symptomology in NPH. The results confirm our earlier hypothesis that reduced caudate volume is associated with increased apathy, and may implicate the striatal regions in cognitive and motivational impairments experienced by patients with NPH.

ACK N OWLED G EM ENTS
The authors would like to thank P. Simon Jones for his help and advice during the revision of this manuscript. The authors would also like to thank the administrative staff at the Departments of Psychiatry and Clinical Neurosciences, University of Cambridge, Addenbrooke's Hospital for their assistance.