Uniformly positive or negative correlation of cerebral gray matter regions with driving safety behaviors of healthy older drivers

This study investigated the relationship between cerebral gray matter (GM) regions and driving safety behaviors (DSBs) of 98 older drivers without dementia (mean age, 77.72 ± 3.677 years). Their DSBs were evaluated on actual vehicles running on a closed-circuit course. The DSB was scored in six categories: DSB1, visual search behavior; DSB2, speeding; DSB3, signaling of the indicator; DSB4, vehicle stability; DSB5, positioning; and DSB6, steering. The scores were calculated by a single driving instructor; larger scores indicated safer driving performances. Regional GM volumes were measured with voxel-based morphometry by magnetic resonance imaging (MRI). Out of 56 GM regions, 18 were correlated with DSB categories except for DSB4. When a single GM region was correlated with multiple DSB categories, a positive or negative response was uniformly determined for the respective region despite clear differences in the DSB categories. This result suggests the possible existence of two contradictory mechanisms in the brain for DSB. The left postcentral gyrus may largely function in regulating DSBs because it was negatively correlated with five of six DSB categories. Thus, MRI’s measurement of regional GM volumes may help deepen the understanding of the diversity and complexity inherent in brain functions for DSBs.


Measurement of regional brain volumes
T1-weighted MR images were obtained using the 1.5-Tesla ECHELON Vega system (Hitachi, Tokyo, Japan) with the three-dimensional gradient echo with an inversion recovery sequence.The following scanning parameters were used: repetition time, 9.2 ms; echo time, 4.0 ms; inversion time, 1000 ms; flip angle, 8°; field of view, 240 mm; matrix size, 0.9375 × 0.9375 mm; slice thickness, 1.2 mm; and the number of excitations, 1.Each image was visually assessed for brain diseases and anomalies, head motion, and artifacts affecting the volumetric measurement.We used the VBM8 toolbox (http:// dbm.neuro.uni-jena.de/ vbm/), which is specially designed for VBM analysis, and other modules implemented in Statistical Parametric Mapping (SPM) 8 (https:// www.fil.ion.ucl.ac.uk/ spm/) to estimate regional brain volumes.In the toolbox, Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) and a custom-made template using IXI Dataset (https:// braindevel opment.org/ ixi-datas et/) were used.The DARTEL algorithm has been shown to sensitively detect atrophy in the brain of patients with Alzheimer's disease 18 .
In brief, the images were segmented into GM, WM, and cerebrospinal fluid space using the maximum a posteriori (MAP) approach 19,20 .The segmented GM and WM images were then used to estimate the morphological correspondence between the template image and the participant's brain using the high-dimensional nonlinear warping algorithm 21 .The estimated nonlinear warp was inversely applied to an atlas defined in the template space to parcellate the target brain anatomically.The neuromorphometrics atlas incorporated in SPM12 was used for the parcellation, with a modification for WM lesions, which appeared as incorrect GM segments around the lateral ventricles.The volumes of 56 anatomical regions were calculated as the sum of the correspondent tissue densities in the voxels belonging to each region.

Evaluation by DSBs
Actual vehicle driving experiments were performed on a closed-circuit course (Fig. 1a), officially designated for renewing drivers' licenses for older drivers by the National Police Agency (The Driver's License Skill Test Implementation Standard), in the Aki Driving School in the Chugei area, Kochi, Japan (Fig. 1a).In the present test, six locations on the driving course were selected for rating.These locations included changing lines when driving straight (Fig. 1b, P1), changing lanes when driving straight; P2, intersection with one right turn; P3, straight course; P4, intersection with one left turn; P5, large curve with poor visibility; P6, another right turn having a stop sign.
A Toyota four-wheel, 1400-cc vehicle (COMFORT) was used.The typical speed of the vehicles on the closedcircuit course ranges from 20 to 50 km/h, and approximately 20 min is taken to complete a circuit.As a good sample of DSB, the same official driving instructor completed the assessment after teaching all participants how to drive.No other driving events were included in the test.In particular, a single instructor drove the course and demonstrated good driving performance once, while participants sat next to him.Then, the participants drove the course once with the instructor sitting on the passenger seat.The instructor assessed each participant's driving skills at six locations with a 3-point scale based on six category methods described previously 12,13 .They responded to the items using a three-point scale: 1, poorly done; 2, normally done; and 3, well done.These rating scores at six locations were then calculated as the "overall evaluation" by assessing the six categories: "visual search behavior (safety recognition with head movement), " "speeding (choice of vehicle speed), " "signaling (timely and appropriate usage of the indicator), " "vehicle stability (acceleration and braking without knocking and completely pulling up in front of the stop line), " "positioning (vehicle movement along the radius of the curvature at intersections without large or small turns), " and "steering (smooth handling with appropriate starting and ending)." Larger scores indicated stronger compliance with the Road Traffic Act.An average value of the summed scores at the six locations for the two rounds of the course was calculated for the DSBs.

Statistical analysis
Multivariate linear regression analysis using IBM SPSS Statistics was employed to estimate the impact of various conditions on each DSB category.The dependent variables in the regression models were the DSB categories, whereas the independent variables included driving distance and specific brain regions.Age and gender were included as covariates to account for their potential influence on DSBs.
In this study, 18 regression models were constructed, with 3 regression models corresponding to each DSB category.The DSB category served as the dependent variable in each model.The independent variables for each of the three regression models were as follows: driving distance (weekly driving distance of < 50 km, 1; between 50 and 200 km, 2; between 200 and 500 km, 3; more than 500 km, 4); age; gender (women, 1; men, 0); and brain conditions.Brain conditions consisted of three models: total brain (Table 1), cerebral lobes (Table 2), and 56 specific GM regions (Table 3).The total brain model consisted of total brain, GM, and WM volumes.The cerebral lobes model consisted of the volumes of the frontal, temporal, parietal, and occipital lobes.The GM regions model consisted of the volumes of 56 specific GM regions.
To determine which cerebral regions are significantly associated with DSBs, we constructed linear regression models for each combination of DSB category and independent variable.Bonferroni's correction was applied to correct for multiple comparisons in these linear regression analyses.
Statistical significance was defined as a p-value of < 0.05.In addition, variance inflation factors (VIFs) were calculated as a part of the collinearity tests conducted alongside the linear regression models to assess the presence of collinearity among the independent variables.This comprehensive approach allowed us to examine the relationships among various brain structures, driving distance, and DSBs while controlling for age and gender.The use of multiple regression models, each tailored to specific sets of independent variables, enabled a more nuanced understanding of the factors influencing DSBs across different DSB categories.
The adjusted alpha value for significance was uniformly calculated across all models based on this approach.Specifically: www.nature.com/scientificreports/For each model, irrespective of the number of independent variables it contained (whether 6 as in Table 1, 7 as in Table 2, or 59 as in Table 3), the Bonferroni-corrected significance level was consistently set at 0.05/18 ≈ 0.0028.This approach ensured a uniform standard for determining statistical significance across all models, thereby mitigating the likelihood of false positives that could arise from multiple testing.Additionally, we have revised our approach to utilize the False Discovery Rate (FDR) method by Benjamini and Hochberg, which better accounts for the interdependencies among multiple variables and provides a more accurate reflection of significant associations in our data.The FDR value is represented as the q-value in Tables 1, 2, and 3.

Ethics statement
This study was conducted in accordance with the "Ethics Guideline for Medical and Health Research Involving Human Subjects" based on the Declaration of Helsinki.All participants signed a formal agreement outlining that the experimental data would only be used for scientific study and that the results would ensure anonymity.Written informed consent was obtained from all participants.This study was approved by the institutional review board at Kochi University of Technology (Application no.C4-3).

Regression with total brain, GM, and WM volumes
We determined the association between DSB scores and independent variables, including total brain volume, GM volume, and WM volume.Six linear regression models were created.DSB1 had an effect size of 0.1417 (p < 0.05), DSB2 had an effect size of 0.0526 (p > 0.05), DSB3 had an effect size of 0.1226 (p > 0.05), DSB4 had an effect size of 0.1032 (p > 0.05), DSB5 had an effect size of 0.1208 (p < 0.05), and DSB6 had an effect size of 0.0681 (p < 0.01).The effect sizes were calculated utilizing Cohen's f 2 formula: , where R 2 is the coefficient of determination as presented in Tables 1, 2, and 3 for each group of independent variables.This formula was consistently applied across all groups of independent variables to determine their respective effect sizes.
The gender parameter showed a significantly negative correlation with the DSB1, DSB5, and DSB6 scores.In the brain regions, total GM was significantly correlated with DSB2.Males exhibited significantly higher DSB scores in these specific DSB categories.All VIF values for the selected independent variables in the linear regressions remained below 2, indicating an absence of multicollinearity among age, gender, GM, and WM.

Regression with four cerebral lobes
We explored the relationship between DSB scores and the volume of the four cerebral lobes.Six linear regression models were created.DSB1 had an effect size of 0.2094 (p < 0.05), DSB2 had an effect size of 0.06 (p > 0.05), DSB3 had an effect size of 0.1287 (p > 0.05), DSB4 had an effect size of 0.1219 (p > 0.05), DSB5 had an effect size of 0.1174 (p > 0.05), and DSB6 had an effect size of 0.0645 (p > 0.05).The results are summarized in Table 2.
The gender parameter demonstrated a significant negative correlation with DSB1, DSB4, DSB5, and DSB6, indicating that males exhibited higher DSB scores in these categories.No statistically significant correlations were observed in the four cerebral lobes.
Importantly, all VIF parameters for the selected independent variables used in the linear regressions were below 10, confirming the absence of severe collinearity among age, sex, frontal, temporal, parietal, and occipital lobes.

Regression with 56 GM regions
Our analysis extended to 56 subdivided GM regions, and the outcomes are summarized in Tables 3 and 4. Six linear regression models were created for the 56 GM regions.DSB1 had an effect size of 2.2349 (p > 0.05), DSB2 had an effect size of 2.6763 (p < 0.05), DSB3 had an effect size of 0.9374 (p > 0.05), DSB4 had an effect size of 1.8775 (p > 0.05), DSB5 had an effect size of 1.299 (p > 0.05), and DSB6 had an effect size of 1.8558 (p > 0.05).
Of the 56 GM regions, based on the VIF value, 18 GM regions exhibited statistically significant correlations with some DSB categories, with the exception of DSB4 (Table 3).In some of the 18 GM regions, the VIF exceeded www.nature.com/scientificreports/10.As a result, we conducted a collinearity assessment using these 18 significantly correlated GM regions, revealing VIF parameters all below 8.This supports the assertion that severe collinearity does not exist among age, sex, and the 18 GM regions, as indicated in Supplemental Table 1.The significantly correlated GM regions are shown in Fig. 2. The correlated parts are colored blue and red for positive and negative correlation, respectively."Positive" means higher volumes related to higher scores in a respective DSB category, whereas "negative" means lower volumes related to higher scores in the respective DSB category.
Notably, the left postcentral gyrus displayed negative correlations with five DSB categories, while the left temporal, left angular, left supramarginal, and right parahippocampal gyrus were positively correlated with two DSB categories.The right angular gyrus was negatively correlated with two DSB categories.Additionally, the right inferior and middle frontal gyrus, precentral gyrus, entorhinal area, left occipital, and fusiform gyrus exhibited positive correlations with one DSB category.The right temporal, lingual, supramarginal gyrus, cuneus, and the left middle frontal and precentral gyrus displayed negative correlations with one DSB category.When a GM region correlated with multiple DSB categories, all correlations were positive or negative, which indicates a uniform determination of positive or negative for a GM region (Table 4).
Crucially, post Bonferroni and FDR corrections, only the left postcentral gyrus retained statistical significance with DSB-1 (visual search behavior) as can be seen in Table 3, emphasizing the necessity to interpret results with caution, particularly concerning the robustness of associations between GM regions and DSB categories.

Uniform determination of positive or negative correlation
Initially, we speculated that the volumes of different GM regions are positively correlated with the scores of different DSB categories, because brain atrophy causes both mental and physical decline resulting in the degradation of safe driving skills, such as visual and spatial cognition, attention, decision-making, and execution ability [21][22][23] .Nevertheless, this study revealed a regularly positive or negative correlation of each GM region with DSBs without exception.Ten of 56 GM regions were positively correlated with DSBs, while 8 GM regions were negatively correlated (Table 4).The negative reaction, which is contrary to the initial speculation, may be due to inefficiency and dedifferentiation often reported in aging brains 24 .It may alternatively be involved in a biological defense reaction.We previously examined the relationship between fatigue perception intensity and regional GM volumes in healthy subjects and found that there are regions where GM volume increases for compensation as fatigue increases 25 .In any case, there is no doubt that the relationship between DSB and GM region is more complex than expected.Furthermore, this finding suggests the possible existence of a brain mechanism for DSBs in which two GM regions with positive or negative correlations are uniformly organized regardless of the categories.If the positive or negative volumetric reactions reflect the hyperactivity or hypoactivity of neuronal functions between GM regions, the neuronal network for DSBs may contain the interaction between positive and negative responses in the complex space of the brain, in other words, a complex of two opposite directions, Table 4. Positive and negative correlations of 18 Gy matter regions with statistical significance (P < 0.05) to driving safety behaviors.Blue dots, positive estimates; Red dots, negative estimates.The intracranial volume for each corresponding participant normalized all volumes.Six DSB scores: DSB1, visual search behavior; DSB2, speeding; DSB3, signaling of the indicator; DSB4, vehicle stability; DSB5, positioning; and DSB6, steering.

Explanatory variable
Objective variable-Driving safety behavior (DSB) DSB1 DSB2 DSB3 DSB4 DSB5 DSB6 Gray matter regions i.e., dual mechanics such as Yin and Yang.In the next step, a functional MRI (fMRI) study is planned to elucidate the neuronal connectivity regulating DSBs 26 .

Gender differences
Considering the overall results in Tables 1 and 2, some DSB categories responded negatively to brain volumes for the female participants compared with the male participants.A difference in the brain structure of males and females may have caused this effect 27 .Gender differences may have a greater impact on the whole brain, GM, WM, and cerebral lobes than on GM regions, because the significance disappeared in the regression analysis of the 56 GM regions.This may be the reason why GM regions are too small to detect gender differences for

Differences in DSB reactions
The number of significantly correlated GM regions varies by DSB categories: 11 GM regions for speeding (DSB2), 8 for visual search behavior (DSB1), 4 for steering (DSB6), 3 for signaling (DSB3), 1 for positioning (DSB5), and none for vehicle stability (DSB4).Thus, positioning and vehicle stability of DSBs, which are naturally essential for safe driving, were extremely low in correlation numbers.In this study, a participant drives a car with an official instructor sitting next to him/her, and the nervous situation may have largely affected DSB5 and DSB4 compared with other categories of DSBs.Most participants carefully drove approximately 20 km/h (data not shown) on the straight course, except for the six checkpoints.This slow speed or careful driving behaviors may especially correlate with none or only one of the GM regions for DSB4 or DSB5, respectively.The advantage of this study is the investigation using six categories of DSBs when driving an actual motor vehicle on a closed circuit.However, the experimental condition is largely different from free driving in residential areas.This difference may have degaussed the variation in the evaluation of DSBs.Further validation with free driving in residential areas is necessary.

Relationship between the GM region and DSBs
The number of significantly correlated DSB categories varies by GM regions: 5 categories for the left postcentral gyrus, 2 for the right angular and parahippocampal gyrus and the left temporal, angular, and supramarginal gyrus, and 1 for 12 GM regions.Thus, the left postcentral gyrus may play an important role in the sufficient enforcement of DSBs because of the higher frequency although it was negatively correlated with DSBs and belongs to the somatosensory cortex unrelated to motor function 28 .Nevertheless, two previous studies have shown completely different results: the study by the Toyota Central Institute reported the supplemental motor area 15 and another by Keio University revealed the four GM regions, namely, the left superior part of the precentral sulcus, left sulcus intermedius primus, right orbital part of the inferior frontal gyrus, and right superior frontal sulcus 16 .Thus far, no studies except for the two are targeting the relationship between regional GM volumes and driving performances.Therefore, research outcomes should be evaluated carefully because the correlated brain regions may largely change depending on the experimental conditions, kinds of motor vehicles, driving on a closed-circuit course or free driving, driving locations, and methods of DSB evaluation.GM gyri multiply correlated to DSB categories are the following; the left postcentral, angular, right parahippocampal, left temporal, and left supramarginal ones.Regardless of girth or left, the postcentral gyrus is a primary somatosensory area and it may be involved in the fine adjustment of steering wheel and pedal operations.The angular gyrus plays an important role in attention, spatial cognition, and semantic information processing 29 .The parahippocampal gyrus is a cortex region in the medial temporal lobe that surrounds the hippocampus and plays an important role in both spatial memory 30 and navigation 31 .The temporal gyrus serves as multimodal brain functions for language, auditory processing, navigation, and comprehension 32 .The supramarginal gyrus is part of the somatosensory association cortex and is involved in motor planning and execution 33 .These descriptions provide plausible explanations for the relationship between GM regions and DSB categories although left and right specificity may affect DSB.

Right-left symmetry
The correlated GM regions with right-left symmetry were temporal, middle frontal, precentral, angular, and supramarginal gyrus (Table 4).Interestingly, right-left symmetry has a combination of positive and negative responses, that is, the middle frontal, precentral, and angular gyrus were positively or negatively correlated in the right or left cerebral hemisphere, respectively.On the other hand, the temporal and supramarginal gyrus were positively or negatively correlated in the left or right cerebral hemispheres, respectively.The results indicated that associated fibers between the right and left hemispheres may regulate the dual mechanics in brain functions for DSBs.

Utilization of DSB and brain data
Noninvasive brain-imaging techniques, such as near-infrared spectroscopy and electroencephalography, can allow measurements of brain activations within actual motor vehicles but remain challenging in the reduction of motion and electric noises 34 .In this study, volumetric data obtained by MRI were used for analyses.Recently, an fMRI study showed that sensorimotor areas increase their activities after changing the direction of a virtual car on a monitor on the bed of an MRI scanner 35 .If fMRI data can be additionally utilized, it will be possible to reproducibly identify brain regions involved in DSBs from both structural and functional data of the brain although MRI cannot measure in real-time on roads.Older drivers who may drive dangerously could be identified with MRI in advance when renewing their driving license soon.Brain atrophy is largely dependent on lifestyles, smoking, drinking alcohol, less exercise, and lack of sleep 13 .The improvement of lifestyles may not only affect the whole brain but also the cerebral regions and lead to the upregulation of driving performances.

Limitations
The interpretation of our study results should be approached with caution due to several inherent limitations.Firstly, the relatively small size of our participant cohort may affect the generalizability of findings.Despite being comparable to or even larger than some prior studies 15,16 , this limitation underscores the need for larger sample sizes in future investigations.
Secondly, the integration of actual driving evaluation and MRI, while offering valuable insights, presents inherent challenges in terms of feasibility.Balancing these two diverse methodologies requires careful consideration, and our study might benefit from advancements in experimental design that address potential confounding factors more effectively.
Thirdly, the controlled environment of the closed-circuit course under the supervision of an instructor, while beneficial for experimental control, may not fully capture the nuances of real-world driving conditions.This discrepancy may influence the assessment of certain DSB categories, particularly positioning and vehicle stability.
Moreover, the absence of a systematic exploration of factors such as sleep, physical condition, and mood during the day represents a limitation.These factors are known to influence DSBs, and their unaccounted variability could impact the study's internal validity.
A critical consideration arises from the significant findings of the relationship between GM regions and DSBs.Post Bonferroni and FDR corrections, only the left postcentral gyrus retained statistical significance with DSB-1 (visual search behavior).This underscores the importance of acknowledging the impact of multiple comparisons on the interpretation of results and raises questions about the robustness of associations between GM regions and DSB categories.
Furthermore, the present study diverged from previous methodologies employed in the field, encompassing variations in locations, point scales for evaluation, and DSB categories.This departure hampers direct comparisons with earlier findings, emphasizing the importance of contextualizing our results within the specific parameters of our study.
Lastly, our participant pool exclusively comprised individuals over the age of 70.While this demographic choice aligns with the focus on older drivers, it raises questions about the generalizability of our findings to a broader age spectrum.To establish the universality of the observed brain-DSB relationships, future studies should encompass diverse age groups.
These limitations collectively highlight the nuanced nature of the brain-DSB relationship, necessitating careful consideration of contextual factors and the iterative refinement of methodologies in future investigations.

Figure 1 .
Figure 1.An actual vehicle and a closed-circuit course.(a) A view from inside the vehicle.(b) Map of the driving course with six rating points.P1, changing line when driving straight; P2, intersection with one right turns; P3, straight course; P4, intersection with one left turn; P5, large curve with poor visibility; P6, another right turn having a stop sign.The corresponding author owns the copyright of the photography.

Table 1 .
Coefficients and corresponding significance levels through regression analyses with gray matter, white matter, and total brain volumes.All volumes were normalized by the intracranial volume for each Significant value after Bonferroni correction.q q < 0.05, the significant value of false discovery rate method. b
DSBs.Further verification using large samples is necessary because of the established gender difference in brain structures.