3.1 Participant Characteristics
The study comprised 171 participants, including 89 CN individuals, 58 patients with MCI, and 24 patients with AD. The CN group was further categorized into two subgroups based on APOE status: 61 APOE-negative (CN-) individuals and 28 APOE-positive (CN+) individuals. Age did not significantly differ among the CN-, CN+, MCI, and AD groups. However, the proportion of females was significantly higher in the CN + group than in the other three groups. Consistent with previous findings, education level significantly differed between the CN- and AD groups, with the latter displaying lower education level. Cognitive function was poorer in the MCI and AD groups when compared to the CN + and CN- groups.
30 CN- participants and 35 patients with MCI or AD were included in the longitudinal analysis. No significant differences were observed between the two groups regarding age, sex, education level, and MRI scanning interval. Regarding the MCI group, participants were further stratified into progressive (n = 11) and stable (n = 22) groups based on subsequent evaluations. There were no significant differences observed between the two groups regarding age, sex and education level (refer to Table 1).
3.2 NODDI-based Analysis of Brain Abnormalities in Cerebral Cortex and Subcortical Structures across AD Spectrum
In the cerebral cortex, individuals in the CN + group showed reduced ICVF in several brain regions compared to the CN- group. These regions include the bilateral lateral temporal lobes (specifically the bankssts, inferior temporal gyrus, and middle temporal gyrus), as well as the bilateral fusiform and left lateral occipital gyrus. Both the MCI and AD groups exhibited widespread abnormalities in brain areas vulnerable to AD. These abnormalities were characterized by decreased ICVF, ISOVF, and altered ODI in specific regions. Specifically, the MCI and AD groups displayed decreased ISOVF and ODI in the extensive lateral temporal, inferior parietal, precuneus, para-hippocampal, posterior cingulate, fusiform, entorhinal and insula cortex alongside increased ISOVF. The insula and fusiform cortex exhibited decreased ICVF, increased ISOVF, and increased ODI (PFDRcorrected<0.05).
Regarding subcortical brain structures, the CN + group demonstrated significantly lower ODI in the right putamen compared to the CN- group. The MCI and AD groups exhibited significantly higher ISOVF and lower ICVF and ODI in both bilateral putamen. Moreover, compared to the CN- group, higher ISOVF, lower ICVF, but higher ODI were observed in the caudate for both the MCI and AD groups (PFDRcorrected<0.05). These findings are presented in Fig. 1. Please refer to supplementary Table 1–3 for the cluster size, peak coordinates, and T statistics of the voxel-wise comparisons of ICVF, ISOVF, and ODI between the CN+-CN-, MCI-CN-, and AD-CN- groups. Additionally, no significant differences were found when analyzing MAP-MRI and DTI indices between CN + and CN- groups.
3.3 Longitudinal Investigation of NODDI Metrics
The longitudinal analysis revealed a strong resemblance between the patterns of NODDI metric changes and the results obtained from the cross-sectional analysis, both in terms of overall trends and specific brain regions. In the MCI group, a significant decrease in ICVF and a notable increase in ISOVF were observed after the follow-up period. Additionally, ODI demonstrated simultaneous increases and decreases across various brain regions. Specifically, in the MCI group, there was a significant decrease in ICVF in the right bankssts, left superior temporal gyrus, and precentral gyrus. Conversely, there was a notable increase in ISOVF in the bilateral thalamus, hippocampus, lateral temporal lobes, and inferior parietal gyrus. Furthermore, ODI exhibited a decrease in the right putamen, left posterior cingulate gyrus, and infratemporal gyrus, while an increase was observed in the bilateral thalamus and lingual gyrus (PFDRcorrected<0.05). These findings are presented in Fig. 2.
Notably, the rate of ICVF increase in the right precuneus, superior parietal gyrus, insula, and left thalamus, fusiform gyrus, and precentral gyrus of the MCI group was significantly higher than that in the CN- group (PFDRcorrected<0.05). Additionally, in patients with MCI and AD, there were significant increases in the rate of ODI in bilateral thalamus, hippocampus, and right superior temporal gyrus, as well as increased ISOVF in the left superior parietal lobe, when compared to the control group. However, it should be noted that these results were not corrected for multiple comparisons (Supplementary figure.1).
Regarding to the MCI subgroup analysis, we observed that the change rate of ISOVF in the left precuneus, putamen, hippocampus and amygdala was significantly higher than that in MCI stable group (Precuneus: SE = 2.17, t = 2.408, P = 0.024; Putamen: SE = 1.74, t = 2.073, P = 0.047; Amygdala: SE = 9.12, t = 2.240, P = 0.033). Additionally, at baseline, the ISOVF in the posterior cingulate gyrus and medial prefrontal cortex in the MCI progressive group was significantly higher than in the MCI stable group (P<0.05). However, the change rate and baseline ICVF and ODI were not significantly different from those of MCI stable group (Figure.2).
3.4 Image correlations between PET and NODDI
In the MCI group, significant positive correlations were observed between ISOVF and tau in various brain regions, including the subcortical regions of the hippocampus, amygdala, and nucleus accumbens, as well as the cortical regions of banksst, cuneus, entorhinal, fusiform, inferior parietal, inferior temporal, isthmus cingulate, lateral occipital, medial orbitofrontal, middle temporal, para-hippocampal, posterior cingulate, and insula cortex. Additionally, there was a significant correlation found between increased ISOVF in the hippocampus, caudal anterior cingulate cortex, inferior temporal cortex, and middle temporal cortex with Aβ (PFDR corrected<0.05, Fig. 3, Supplementary Table 4). It is worth noting that these brain regions mainly belong to the AD vulnerable areas. However, no significant correlations were found between ICVF, ODI, ISOVF, and Tau, Aβ in the CN+, CN-, and AD groups, after FDR correction for multiple comparisons.
3.5 Correlations with neuropsychological measurements
In the group with MCI, we observed that lower ICVF in the right caudate and lower ODI in the left caudate, posterior cingulate gyrus, right putamen, insula, and thalamus were positively associated with impaired delayed memory performance. Moreover, higher ODI in the bilateral thalamus and ISOVF in the right caudate, bilateral superior temporal gyrus, right hippocampus, para-hippocampal gyrus, posterior cingulate gyrus, frontal gyrus, and left fusiform gyrus were also positively correlated with poor delayed memory performance (PFDR−corrected<0.05). ISOVF in the bilateral superior temporal gyrus and left fusiform gyrus showed positive correlations with ADAS scores (PFDR−corrected<0.05). Table.3 and Supplementary Table.5 provides detailed ROI locations, corrected p-values, and correlation coefficients.
In the CN- group, higher ISOVF in the right caudate, posterior cingulate gyrus, frontal gyrus, superior temporal gyrus, and left fusiform gyrus were positively associated with impaired delayed memory performance. Higher ODI in the right insula was positively correlated with better delayed memory performance (PFDR−corrected<0.05). These associations between NODDI metrics and cognitive performance were not observed in the CN + and AD group.
3.6 Mediation analysis
The mediation analysis using bootstrapping revealed significant mediation effects of Tau on delayed memory performance through the mediator ISOVF in the entorhinal, fusiform, inferior temporal, middle temporal, para-hippocampal, and posterior cingulate cortex. Specifically, in the entorhinal cortex, the indirect effect was 0.251 (95% CI [-4.1352, -0.1596]); in the fusiform cortex, the indirect effect was 0.645 (95% CI [-2.9064, -0.8770]); in the inferior temporal cortex, the indirect effect was 0.328 (95% CI [-2.2283, -0.4624]); in the middle temporal cortex, the indirect effect was 0.381 (95% CI [-2.9190, -0.5489]); in the para-hippocampal cortex, the indirect effect was 0.297 (95% CI [-3.2587, -0.2145]); and in the posterior cingulate cortex, the indirect effect was 0.250 (95% CI [-3.2210, -0.3109]).
The proportion of the total effect attributed to the indirect effect was estimated to be 25.1%, 64.5%, 32.8%, 38.1%, and 25% for the entorhinal, fusiform, inferior temporal, middle temporal, and para-hippocampal cortex, respectively (calculated as the indirect effect divided by the total effect) (Fig. 4, Supplementary Table.6).