Shape analysis of the amygdala, hippocampus and thalamus in former American football players

Abstract

Repetitive head impacts are common in contact and collision sports and are linked to structural brain changes and an elevated risk of neurodegenerative diseases such as Chronic Traumatic Encephalopathy. Identifying early in vivo structural markers remains challenging. Although diagnosis currently requires post-mortem confirmation, clinical symptoms, including cognitive impairment and behavioural changes, are reflected in the diagnosis of Traumatic Encephalopathy Syndrome. These symptoms align with dysfunction in key brain regions-amygdala, hippocampus and thalamus-which support memory, emotion and behaviour and commonly show tau pathology in Chronic Traumatic Encephalopathy. This study uses shape analysis to examine structural differences in these regions between former American football players and unexposed asymptomatic controls and evaluates the influence of age, head impact exposure and clinical diagnosis on brain structure. We analyzed brain morphology in former American football players (n = 163) and unexposed, asymptomatic controls (n = 53). Structural segmentation was performed with FreeSurfer 7.1, and the shape analysis pipeline was used to generate subregional reconstructions. Vertex-level morphometry, based on the logarithm of the Jacobian determinant and radial distance, quantified local surface area dilation and thickness. Group differences were examined with covariate-adjusted linear regression models contrasting football players and controls, as well as participants with and without a Traumatic Encephalopathy Syndrome diagnosis. Partial correlations examined the influence of age, age of first football exposure and cumulative head impact index metrics, including frequency, linear acceleration and rotational force. Models were adjusted accordingly for age, body mass index, education, race, imaging site, apolipoprotein $ϵ4$ status and total intracranial volume. Former football players exhibited bilateral surface area contractions in the hippocampus and amygdala, along with reduced amygdala thickness, compared to controls. Older age was associated with widespread surface contractions and thinning across all regions, except for preserved thickness in the left hippocampus. An earlier age of first exposure to football correlated with surface contractions in the thalamus and left hippocampus. Greater cumulative linear acceleration was linked to bilateral hippocampal surface contractions and reduced thickness in the left thalamus, while greater rotational force exposure was associated with hippocampal thinning. No significant structural differences were found between players with and without a diagnosis of Traumatic Encephalopathy Syndrome. These findings extend volume-based research by revealing localized alterations in surface area dilation and thickness and emphasize the roles of age and repetitive head impact exposure in long-term brain changes.

Keywords: neuroimaging; repetitive head impacts; shape analysis; sports-related head injury; structural MRI.

Graphical Abstract

Figure 1

Structural alterations associated with RHI exposure. Linear regression analyses demonstrating surface area contractions and thickness reductions in former American football players (N = 157) compared to unexposed asymptomatic controls (N = 51). (A) As assessed by the Jacobian determinant, surface area contractions were revealed via negative values in the amygdala and hippocampus. In the amygdala, left-side contractions were observed in the basolateral, basomedial, centromedial and lateral nuclei (Cohen’s d = −0.101, P = 0.02), while right-side contractions were found in the basomedial and lateral nuclei (Cohen’s d = −0.103, P < 0.001). In the hippocampus, left-side contractions were noted in the CA1, CA2 and subiculum subregions (Cohen’s d = −0.112, P = 0.001), while right-side contractions were observed in the CA1, CA2, CA3/DG and subiculum (Cohen’s d = −0.084, P = 0.01). (B) Thickness reductions, as assessed by radial distance, showed negative values in the amygdala. Left-side reductions were observed in the basolateral, basomedial and lateral nuclei (Cohen’s d = −0.419, P = 0.001), while right-side reductions were found in the basolateral, basomedial, centromedial and lateral nuclei (Cohen’s d = −0.269, P = 0.01). DG, dentate gyrus.

Figure 2

Age-related structural decline in the amygdala, hippocampus and thalamus among former football players. Partial correlation analyses demonstrating interaction with age in former American football players (N = 163). (A) Jacobian determinant values revealed significant age-related surface area contractions in the amygdala, hippocampus and thalamus. In the amygdala, left-side contractions were observed in the basolateral, basomedial, centromedial and lateral nuclei (Cohen’s d = −0.007, P = 0.04), with similar right-side contractions (Cohen’s d = −0.006, P = 0.04). In the hippocampus, left-side contractions were noted in the CA1, CA2, CA3/DG and subiculum subregions (Cohen’s d = −0.008, P = 0.02), and right-side contractions were observed in the same subregions (Cohen’s d = −0.007, P = 0.03). In the thalamus, left-side contractions were found in the anteroventral, ventral, medial-lateral and pulvinar subregions (Cohen’s d = −0.004, P = 0.04), while right-side contractions were restricted to the anteroventral, ventral, medial-lateral and pulvinar subregions (Cohen’s d = −0.004, P = 0.04). (B) In the analysis of radial distance in the amygdala, left-sided reductions were observed in the basolateral, basomedial, centromedial and lateral nuclei (Cohen’s d = −0.019, P = 0.03), and right-sided reductions were seen in the same nuclei (Cohen’s d = −0.016, P = 0.04). In the hippocampus, right-sided reductions were noted in the CA1, CA2, CA3/DG and subiculum subregions (Cohen’s d = −0.022, P = 0.03). In the thalamus, left-sided reductions were observed in the anteroventral, ventral, medial, lateral, and pulvinar subregions (Cohen’s d = −0.021, P = 0.04), and right-sided reductions were seen in the same subregions (Cohen’s d = −0.018, P = 0.03). DG, dentate gyrus.

Figure 3

Age-related thalamic changes in unexposed controls. Partial correlation analyses demonstrating interaction with age in unexposed asymptomatic controls (N = 53). (A) Jacobian determinant values revealed significant interactions with age in the thalamus bilaterally, indicating surface area contractions, specifically in the ventral and pulvinar subregions (left: Cohen’s d = −0.006, P < 0.0001; right: Cohen’s d = −0.008, P < 0.01). (B) An interaction with age was also observed for radial distance values in the right thalamus, specifically in the ventral, medial, lateral and pulvinar subregions (Cohen’s d = −0.031, P < 0.01), indicating an association between older age and reduced thickness.

Figure 4

Age of first football exposure relates to hippocampal and thalamic morphology. Partial correlation analyses demonstrating interaction with the age of first exposure to tackle football (N = 163). (A) Jacobian determinant values revealed significant interactions with the age of first exposure to tackle football in the left hippocampus, showing surface area expansion in the CA2 subregion (Cohen’s d = 0.026, P < 0.001). In the thalamus, surface area expansion was observed in the left anteroventral, ventral, and medial subregions (Cohen’s d = 0.009, P < 0.01), as well as the right ventral subregion (Cohen’s d = 0.010, P < 0.001). (B) An interaction with age of first exposure was also observed for radial distance values, with increased thickness noted in the ventral subregion of the left thalamus (Cohen’s d = 0.04, P < 0.0001). For clarity, positive coefficients reflect expansion with higher age at first exposure; the inverse holds for earlier exposure (contraction).

Figure 5

Cumulative linear acceleration burden is linked to hippocampal and thalamic morphometric loss. Partial correlation analyses demonstrating interaction with CHII linear acceleration (N = 163). (A) Jacobian determinant values showed significant interactions with CHII linear acceleration in the hippocampus, indicating surface area contraction. Left-side contractions were observed in the CA1, CA2, CA3/DG and subiculum subregions (Cohen’s d = −1.22E-06, P < 0.01), while right-side contractions were found in the same subregions (Cohen’s d = −5.61E-07, P = 0.03). (B) Radial distance analysis revealed an interaction with linear acceleration, showing reductions in thickness bilaterally in the hippocampus (CA1, CA2, CA3/DG, and subiculum subregions (left: Cohen’s d = −2.49E-06, P = 0.03; right: Cohen’s d = −2.19E-06, P = 0.01), as well as in the left thalamus, specifically in the medial and pulvinar subregions (Cohen’s d = −2.18E-06, P < 0.001). CHII, cumulative head impact index; DG, dentate gyrus.

Figure 6

Cumulative rotational force burden predicts hippocampal thinning. Partial correlation analyses demonstrating interaction with CHII rotational force (N = 163). Analysis of radial distance revealed an interaction with CHII rotational force, showing an association between increased rotational force and reduced cortical thickness in the hippocampus. Specifically, left-side reductions in thickness were observed in the CA1, CA2, CA3/DG, and subiculum subregions (Cohen’s d = −4.65E−08, P < 0.01), while right-side reductions were noted in the CA1, CA2 and subiculum subregions (Cohen’s d = −2.72E−08, P < 0.01). CHII, cumulative head impact index; DG, dentate gyrus.