The human amygdala: a systematic review and meta-analysis of volumetric magnetic resonance imaging

Abstract

The structure and function of the human amygdala is attracting increasing attention in the scientific literature, particularly since the advent of high resolution magnetic resonance imaging (MRI). We carried out a systematic review of the published literature reporting left and right amygdala volumes from MRI in non-clinical subjects. Our aim was to estimate the normal range of the volume of the amygdala and to account for heterogeneity of the measures. The factors we considered included the detail given regarding various subject factors, the plane of scan acquisition, slice thickness and contiguity, magnet strength, positional and volume correction, and the reliability of measurement. Thirty-nine studies with 51 data sets fulfilled selection criteria. The mean±95% confidence interval for the left amygdala volume was 1726.7 mm³±35.1, and right was 1691.7 mm³±37.2. The left-right difference did not reach statistical significance. The overall range of reported volumes was 1050 mm³ to 3880 mm³. The amygdala is significantly larger in men and shows an inverse correlation with age. The main methodological factor found to influence amygdala measurement was anatomical definition. Studies using ‘Watson’s criteria’ (Neurology 42 (1992) 1743) produced significantly larger volumes than the remainder. An index of study quality revealed an inverse relationship with volume—the higher the quality the smaller the volume. This reflected such factors as slice thickness, correction for brain volume, positional correction and number of subjects. We conclude by putting forward a detailed operationalised anatomical delineation of the amygdala, based on Watson’s criteria. This work should guide future research in obtaining accurate and reliable amygdala volume measures which in turn will aid comparisons with clinical groups and the specification of structural–functional relationships.

Introduction

The amygdala is a bilateral, ovoid, grey matter structure, on the superomedial surface of the temporal lobe, anterior to and partially overlapping with the hippocampal formation [24]. It is composed of several cortical and subcortical nuclei [70]. Experimental investigations of clinical and normal populations, many using non-invasive neuroimaging techniques have identified the role of the amygdala in emotional memory [15], [16], emotional facial expression recognition [3], [81], emotional auditory recognition [66] and fear conditioning [41].

The resolution of structural magnetic resonance imaging (MRI) is sufficient to enable quantitative assessment of the volume of the amygdala complex, although the individual nuclei that comprise this heterogeneous structure are not distinguishable. For example, it is not at present possible to measure the volume of the cortical or lateral amygdala. Moreover, because the amygdala borders are not all clear owing to proximity to other grey structures, most methodologies rely on external landmarks and stereotactic methods to improve reliability. For instance, the anterior border, which can be especially difficult to define, may be defined arbitrarily at the slice which shows the lateral sulcus closing to form the endorhinal sulcus [74] or alternatively at the level of the mamillary bodies [2]. Both methods necessarily mean that individual anatomical variation may influence measurement, but the border is unambiguous. Alternatively, the most anterior slice has been identified less precisely as the point where the amygdala ‘no longer has an ovoid shape’ [71], [12]. Hence definitions of the normal amygdala are prone to differences in interpretation and methodology.

MRI has been used to infer pathology or maldevelopment of the amygdala in vivo, in diverse clinical populations [71], [68], [29], [37]. For example, individual studies have found that Alzheimer’s patients show signs of 33% amygdala atrophy [23], male schizophrenic patients exhibit amygdala volume reductions of 8%, contrasting with a 10.5% increase in female amygdala volumes [30], and amygdala volumes in one fifth of a series of chronic temporal lobe epilepsy patients were 20% smaller than normal [36]. Increases in amygdala volume have also been claimed in autism [33] and bipolar affective disorder [4], [5]. Much of this ever-growing literature is contradictory or inconclusive [21]. Moreover, inferences about pathology based on volume reductions or increases require reliable norms and variance estimates.

Meta-analysis of structural neuroimaging research has been valuable in summarising similar data [78], [79], [67]. Hence we carried out a systematic review of the published scientific literature on volumetric MRI of the human amygdala in non-clinical subjects in order to determine the volume of the ‘normal’ amygdala and its upper and lower limits, using standard meta-analytic techniques. We were particularly interested in asymmetry of the amygdala. Lateralised differences in the functioning of the hippocampus have long been recognized [50] while analogous differences in amygdala function have only recently attracted interest. This has been motivated by functional neuroimaging which has shown asymmetrical activation in response to fearful or generally emotive stimuli [53], [64], [34], [28], [58]. This in turn has been explained by the degree of conscious awareness [54], [77], the valance of the emotion [18], the speed of habituation [14], [59] and particularly, interactive effects between laterality and sex differences [17], [38]. The question of structural asymmetries in amygdala volume has not been addressed systematically although information on this issue could provide clues regarding function.

Alongside our aim to determine whether there are reliable differences in the volumes of the amygdala in the right and left cerebral hemispheres, we sought to examine differences between men and women, and those according to age. We also set out to compare anatomical definitions of the amygdala and the effects of MRI scanning parameters on volume measurements such as slice thickness, the use of one or more planes, and volume correction, and to test whether such factors interacted with reported structural asymmetries. Finally, we attempted to assess study quality, in particular study size, and examined its influence on volumetric measures.

The establishment of the normal range of amygdala volumes will have clinical as well as research utility. Clinical radiologists often rely on being able to assess the presumed abnormal amygdala relative to its normal opposite side. However, bilateral disease is common [19], [20] and the possibility of pathological increases as well as decreases obviously undermines the validity of such comparative judgements. From a research perspective, although ‘cases’ can be deemed abnormal in direct comparison to controls assessed using identical measures, generalisability and comparison with other studies is problematic if absolute values differ markedly from study to study. Exposing the causes of heterogeneity in measurement of the amygdala will enable researchers to make more informed choices as to the volumetric methods they employ and will facilitate the accretion of a coherent and interpretable body of knowledge.