Descriptive statistics and sex differences in demographic characteristics, total intracranial volume, and EM performance

As shown in Table 1, compared to females, males had significantly more years of education (t = 2.71, p = 0.007), larger total intracranial volume (TIV) (t = 25.01, p < 0.001), higher SEM scores (t = 3.69, p < 0.001), but lower VEM scores (t=-4.40, p < 0.001). There was no significant sex difference in age (t = 0.27, p = 0.785).

Table 1 Descriptive statistics and sex differencesDifferences in VEM and SEM between age groups

Multiple linear regression models were used to examine the relationship between age and EM after controlling for years of education (as well as sex in the analysis of the total sample). Results showed that both VEM and SEM were negatively correlated significantly with age for the total sample (VEM: β=-0.04, t=-8.72, df = 922, p < 0.001; and SEM: β=-0.03, t=-6.56, df = 922, p < 0.001), for males (VEM: β=-0.04, t=-5.80, df = 460, p < 0.001; and SEM: (β=-0.02, t=-3.44, df=,460 p < 0.001), and for females (VEM: β=-0.04, t=-6.34, df = 460, p < 0.001; and SEM: β=-0.04, t=-5.62, df = 460, p < 0.001). Direct comparisons of linear slopes of memory as a function of age between males and females via the inclusion of interaction terms (sex×age) showed no significant sex difference in age effect for either SEM or VEM (interaction terms p > 0.05).s.

The relationship between GMV and two types of EM

Multiple linear regression model was utilized to investigate the relationship between GMV and EM performance, with results corrected for false discovery rates (FDR = 0.05). The covariates included age, years of education, and TIV (sex was used solely for the analysis of the entire sample).

VEM

The analysis of the total sample found significant correlations of GMV and VEM in a wide range of cortical and subcortical regions, primarily located in the bilateral hippocampus, para-hippocampus, bilateral anterior cingulate gyrus, bilateral insula, bilateral thalamus, and bilateral amygdala. The parahippocampal gyrus showed the strongest correlation (Fig. 1, 1st row; Table S1 in the supplementary material). When analyzed separately for males and females, the results for males were similar to those of the total sample, except that the left amygdala showed the strongest effect (Figs. 1 and 2nd row; Table S2 in the supplementary material). For females, however, there were fewer significant regions, and the strongest effect was found in the temporal pole and middle temporal gyrus (Figs. 1 and 3rd row; Table S3 in the supplementary material).

Fig. 1

The brain regions significantly associated with two types of EM Only the areas with significant positive results are shown. The color bar shows the range of the regression coefficient β, with red indicating the strongest correlation and yellow indicating the weakest effect.

SEM

For the total sample, we found significant correlations of GMV and SEM in a wide range of cortical and subcortical areas dominated by the medial temporal and frontal lobes, such as the bilateral hippocampus, para-hippocampus, bilateral anterior cingulate gyrus, bilateral insula, bilateral thalamus, bilateral amygdala, and a few parietal and occipital regions. The strongest correlation was found in the left hippocampus (Figs. 1 and 4th row, Table S4 in the supplementary material). The analysis of the males found relevant regions similar to those found for the total sample, with the strongest effect in the left hippocampus (Figs. 1 and 5th row, Table S5 in the supplementary material). Females again showed fewer and weaker effects than the males, and the strongest effect was observed in the temporal pole and superior temporal gyrus (Figs. 1 and 6th row, Table S6 in the supplementary material).

Age differences in the relationship between GMV and two types of EM

The sliding windows approach was used to investigate age differences in the relationship between GMV and EM. Our first step was to generate binarized masks of the brain regions that showed significant relationship with each type of EM (refer to results 2.3 for details). The next step involves conducting multiple linear regressions for the group normalized residuals (referring to the residuals of GMV after regressing out covariates, standardized as z-scores across all samples) of GMV (controlling for age, education years, TIV, excluding sex only for all subject group) using two different EM within all windows, respectively. Subsequently, the average t value of the t-test statistic for the regression coefficients β (p < 0.01, uncorrected) within each region from AAL90 served as a measure of the association between EM and GMV within the AAL90 region [39]. The mean t-value within a region of each window was then correlated (Pearson’s correlation) with average age of each window, FDR = 0.05 corrected. Fisher’s z-Tests were then utilized to assess the significance of differences in the correlation coefficients (r) between the two sexes.

VEM

For the total sample, VEM was positively associated with GMV in large areas of the frontal and temporal lobes, as well as in smaller areas of the parietal, occipital, and subcortical lobes (see Fig. 2a). but those involving left superior parietal gyrus, right supramarginal gyrus, left angular, left inferior and middle occipital gyrus, and right putamen were negatively correlated with age (also see Table S7 in the supplementary material).

Fig. 2

Age differences in the correlation between GMV and two types of EM. (a) Age differences in the t values indexing the relationship between GMV and two types of EM for the total sample. (b) A heatmap of the correlations between age and the strength of the GMV- EM association across five clusters of 90 brain regions. Males (outer circle) and females (inner circle) were analyzed separately. Each color block represents a brain region, with yellow blocks indicating nonsignificant regions, red blocks indicating regions with positive correlations, and blue blocks indicating regions with negative correlations. The names of the brain regions are displayed in the legend on the outermost circle. The innermost asterisks indicate the significance of differences in correlation coefficients (* < 0.05, ** < 0.01, *** < 0.001). Only regions where GMV-EM correlations with age were significant in either the male or female group, or in both groups, are displayed. In regions where there are no significant results for GMV-EM, the significance of differences cannot be calculated (for example, in the female group in SEM, the parietal and occipital lobes).

When analyzed by sex, somewhat different results were found by type of EM and sex. Results are summarized in heatmaps (see Fig. 2b), a different manner of presentation than that for the total sample. For males, the strength of the GMV-VEM association positively correlated with age in regions dominated by frontal, temporal, and parietal lobes, with some occipital and subcortical regions, but were negatively correlated with age in certain regions such as right rectus, right supplementary motor area, left postcentral, left inferior parietal gyrus, and left inferior occipital gyrus (Fig. 2b left outermost circle and Table S7 in the supplementary material). For females, positive correlations between the GMV-VEM association and age were found in extensive areas dominated by frontal, temporal, and parietal lobes, as well as some occipital and subcortical areas, but negative correlations were found in the left middle occipital gyrus and right middle temporal gyrus (Fig. 2b left innermost circle and Table S7 in the supplementary material). The results of the comparison of correlation coefficients in the two sex groups are presented in the left portion of Fig. 2b and Table S7 in the supplementary material.

SEM

For the total sample, SEM was positively associated with GMV in extensive areas dominated by frontal, temporal, parietal, and subcortical regions, as well as a small number of occipital regions, and the strength of these associations positively correlated with age (Table S8 in the supplementary material).

When analyzed by sex, we found that in males, the strength of the GMV-SEM association positively correlated with age in large areas of temporal and frontal lobes, as well as some parietal, occipital, and subcortical lobes, but negatively correlated with age in the right inferior occipital gyrus (Fig. 2b right outermost circle and Table S8 in the supplementary material). For females, we found strong positive correlations between age and the strength of the GMV-SEM association in various regions, mainly in the temporal lobe, and some weak positive correlations in frontal, occipital, and subcortical lobe. No region showed a negative correlation in females. The results of the comparison of correlation coefficients in the two sex groups are presented in the right portion of Fig. 2b and Table S7 in the supplementary material.

Analysis at the level of brain systems: the AT and PM systems

Finally, we analyzed the data at the level of brain systems—the AT and PM systems, as discussed in the Introduction. In light of the t values calculated for each window as outlined in Sect. 2.4, the average t values were extracted for each type of EM within the AT and PM systems. Subsequently, Pearson correlation analyses were conducted to assess the relationship between average age of each window and average t values within each system of each window, both within the entire sample and separately for males and females. Fisher’s z-Tests were then utilized to assess the significance of differences in the correlation coefficients (r) between the two sexes.

VEM

For the total sample, age was positively correlated with the strength of the association between VEM and the GMV of the AT system (r = 0.70, p < 0.001), but not that between VEM and the GMV of the PM system (r = 0.11, p = 0.65) (Fig. 3b upper, left). In the analysis based on sex, distinct patterns emerged. Among males, the strength of VEM’s association with GMV in the AT system (r = 0.54, p = 0.011) and the PM system (r = 0.48, p = 0.029) both showed significant positive correlations with age. (Fig. 3b upper, middle). In females, however, age was significantly and positively associated with the strength of VEM’s association with GMV of the AT system (r = 0.71, p < 0.001), but not significantly associated with GMV of the PM system (r = -0.36, p = 0.111) (Fig. 3b upper, right). The sex differences in the correlation coefficient (r) for both systems are highly significant, as evidenced by the following statistical results: AT (Fisher’s Z = -4.29, p < 0.001) and PM (Fisher’s Z = 13.65, p < 0.001).

Fig. 3

Age differences in VEM’s and SEM’s association with GMV of the two brain systems (a) Graphical representation of the AT and PM systems, with red representing the AT system and blue representing the PM system. (b) The age-related differences in VEM and SEM in relation to GMV in two brain systems, with each point in the figure representing the average t value within either the AT system (red) or the PM system (blue) for each window.

Fig. 4

Age distribution in sliding windows for the total sample (a), males (b), and females (c)

SEM

For the total sample, age was positively associated with the strength of SEM’s association with GMV of both the AT (r = 0.79, p < 0.001) and PM (r = 0.56, p = 0.008) systems (Fig. 3b below, left). When analyzed by sex, different patterns were observed. Similar to VEM, in males, the strength of the association between VEM and GMV in both the AT system (r = 0.48, p = 0.028) and the PM system (r = 0.49, p = 0.025) showed significant positive correlations with age (Fig. 3b below, middle). Females showed a strong positive correlation between age and the strength of SEM’s association with GMV of the AT system (r = 0.91, p < 0.001). There was no relevant finding for the PM system because no region within the PM system was significantly associated with females’ SEM scores (Fig. 3b below, right). The sex difference in the correlation coefficient (r) of the AT system is significant (Fisher’s Z = -9.71, p < 0.001). In contrast, the PM system cannot be compared due to the lack of statistical significance in the average t values for females across all windows.