We identified a dietary pattern associated with a lower prevalence of dementia and smaller volumes of WML in older Japanese people. This dietary pattern was characterized by high intakes of green and other vegetables, algae, potatoes, fish, fruits and fruit juices, egg, meat, pickles, soybeans and soybean products, milk and dairy products, and low intake of alcoholic drinks, sugar and confectioneries, roughly corresponding to a Japanese diet including protein and minerals. These findings highlight a dietary pattern that shows statistical associations with both dementia status and brain structural abnormalities in aging, contributing to the understanding of how diet may relate to cognitive and brain health in later life.

DP1 includes dietary components beneficial and potentially harmful to brain health. Vegetables, algae, soybeans, potatoes, and fruits contain dietary fiber, beneficial to brain health via gut-brain interactions involving microbiota [19]. Several studies have shown that soluble high-fiber foods induce neuroinflammation and dementia risk [20,21,22], consistent with our findings. Pickled foods may also positively influence gut microbiota and dementia risk [23]. Fish, seaweed, vegetables, and fruits provide beneficial nutrients, including n-3 polyunsaturated fatty acids, docosahexaenoic acid (DHA), eicosapentaenoic acid (EPA), folic acid, vitamin E, and antioxidants such as polyphenols and carotenoids. Eggs, meat, and dairy products have been reported to support cognitive function [24, 25], possibly due to their cholesterol content, essential for synapse formation and neuronal signaling [26]. Additionally, lutein, choline, zeaxanthin, and protein (abundant in eggs) effectively protect nerves, thereby preventing dementia [27]. Conversely, excessive alcohol intake is associated with dementia due to neurotoxic effects, cardiovascular complications, and thiamine deficiency [28]. High sugar consumption rapidly elevates blood glucose and insulin, potentially impairing brain function [29]. Given that excessive intake of cholesterol, pickles and fruit juices may be associated with a high risk of diabetes and cardiovascular disease, further research is needed to determine the optimal intake.

Several studies have reported that meat, fish, potatoes, vegetables, legumes, fruits, dairy products, and eggs, containing B vitamins, vitamin E, carotenoids, flavonoids, and long-chain omega-3 PUFAS, are associated with greater brain volume and lower WML volumes [30, 31]. These are consistent with our results. Conversely, inadequate intake of B vitamins (folate, B12, B6) increases the risk of hyperhomocysteinemia, contributing to white matter damage [32]. In addition, low serum levels of vitamin C (abundant in fruits, vegetables, and potatoes) and vitamin E (abundant in green vegetables) are associated with more severe DWML volumes [33]. In contrast, higher dietary intake of flavonoids (existing in fruits and legumes) was associated with lower WML volumes [34]. Fish is an excellent source of omega-3 fatty acids, particularly eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), which are essential for neuronal integrity and may help reduce WML burden [35]. Meat, eggs, soy, and dairy provide quality protein for neuronal maintenance and repair [36, 37]. In contrast, excessive sugar and alcohol intake may trigger brain inflammation, oxidative stress, and metabolic disorders, leading to WMLs [38, 39].

WML volumes were quantified using a ConvNet model previously developed, which enables high-resolution detection of WMLs and thus enhances the validity of the study findings that a Japanese dietary pattern including minerals and proteins was associated with lower total WML volume.

WMLs were further categorized into PWML, DWML, and IWML regions based on their distinct etiological and clinical relevance. PWMLs are frequently linked to cerebrovascular dysfunction [40]. Thus, the inverse relationship between DP1 scores and PWML volumes may suggest that a healthier diet is related to reduced cerebrovascular burden. Additionally, DWMLs may be connected with metabolic syndrome, such as obesity [41]. Thus, dietary management might mitigate these metabolic risks and play a role in DWML development. Furthermore, IWMLs have been implicated in cognitive decline [42], and our findings suggest that this dietary pattern may be relevant to early neurodegenerative processes.

Importantly, a sensitivity analysis excluding individuals with dementia was conducted to reduce the potential for reverse causation due to prodromal dietary changes. The persistence of inverse associations between DP1 scores and WML volumes in this non-demented subgroup supports the hypothesis that diet is associated with brain health independent of dementia. Although the cross-sectional design limits causal inference, this sensitivity analysis enhances the biological plausibility of the observed associations. Since WMLs are considered early markers of cerebrovascular and neurodegenerative processes, these results suggest that favorable dietary patterns may be associated with better brain health even before the onset of clinical dementia. The consistent associations observed in the non-demented subgroup support the potential role of diet in cognitive and neurological health and help mitigate concerns regarding reverse causation in cross-sectional research.

Although the association between DP1 scores and log-transformed WML volume was relatively small (e.g., β for total WML volume = − 0.03), even minor variations in WML burden may carry clinical implications. Prior evidence has shown that the presence of WMLs is significantly associated with highly elevated risks of stroke, dementia, and all-cause mortality [43]. These associations underscore the clinical relevance of even mild WMLs and support their role as early indicators of future neurological decline.

Moreover, previous studies have consistently demonstrated that relatively low WML burden is associated with subtle but meaningful deterioration in multiple cognitive domains, including memory, processing speed, attention, and executive function [44]. These findings suggest that WML-related cognitive changes may begin early and gradually progress over time.

WMLs are also recognized as core neuroimaging markers of cerebral small vessel disease (CSVD), a major contributor to vascular cognitive impairment and stroke. Early-stage WMLs may reflect incipient microvascular damage that is potentially reversible, emphasizing their importance in early detection and preventive strategies [45].

From a public health perspective, although the individual-level effect may appear modest, the high prevalence of WMLs among older adults implies that even small associations could lead to a significant cumulative burden at the population level. Therefore, even mild WMLs may warrant consideration in clinical practice.

This study addresses several important gaps by focusing on a previously underrepresented population, applying high-resolution WMLs, and evaluating these associations within a specific cultural context.

This study represents the first standardized nationwide investigation in Japan to examine the associations between dietary patterns, dementia prevalence, and brain structural markers such as WMLs. Using a harmonized FFQ and uniform protocols, we analyzed data from 8,938 older adults recruited from eight study sites across Japan, ranging from Tohoku in the north to Kyushu in the south, encompassing diverse climatic and dietary environments.

Building on this nationwide framework, a site-specific analysis was conducted to evaluate the consistency and generalizability of the association between DP1 and dementia across Japan. Although an overall inverse association between DP1 scores and the prevalence of all-cause dementia and AD was observed in most study sites, statistical significance was not consistently achieved. One exception was Hirosaki, where a positive trend was likely due to the extremely small number of dementia cases.

The identification of a consistent dietary pattern (DP1) across regions and the application of subgroup analyses support the robustness and generalizability of this pattern across distinct regional contexts, distinguishing this study from previous investigations that were often limited in scale and scope.

Although the analytical methods used in this study, including PCA and deep learning-based WML quantification, have been applied in population-based studies, few studies have utilized these approaches in large-scale, multi-regional Japanese samples. This study enables the assessment of dietary patterns across diverse regions, improving the stability and internal consistency of the findings in Japan. The cultural specificity of dietary practices in Japan highlights the importance of conducting such population-specific investigations, rather than relying solely on evidence from different cultural and dietary contexts. These results contribute to the growing body of cross-cultural nutritional neuroepidemiology.

Another strength of the study is the introduction of a novel element by incorporating MRI-based neuroimaging markers through the high-resolution quantification of WMLs. While previous studies have linked specific foods or nutrients to brain structure, the interaction between overall dietary patterns and WMLs remains unclear, particularly within the context of Japanese dietary culture. To our knowledge, this is the first study in Japan to apply this imaging method in dietary pattern research. This approach, based on a validated ConvNet model, provides an objective and sensitive indicator of brain aging and allows for the assessment of associations between dietary behaviors and WMLs in older adults.

However, several limitations should be noted. First, a major limitation of this study is its cross-sectional design, which precludes conclusions about causal relationships between dietary patterns and WMLs or dementia. Although we conducted a sensitivity analysis excluding individuals with dementia in the analysis of dietary patterns and WMLs to reduce the risk of reverse causation, it remains possible that early or subclinical cognitive changes may have influenced dietary behaviors. This concern is particularly relevant in aging populations where lifestyle factors may be modified in response to subtle health changes. Therefore, the observed associations should be interpreted with care, and future longitudinal studies are warranted to investigate how adherence to this dietary pattern influences the development and progression of dementia and WML over time.

Second, dietary intake assessed by FFQ may introduce recall bias, although caregiver assistance was provided for participants with cognitive impairment. Last but not least, the study population consisted solely of older Japanese adults, which may limit the generalizability of the findings to populations with different dietary habits, genetic backgrounds, or age ranges.