Putra Handityo Aulia, Park Kaechang, Yamashita Fumio
Department of Electrical, Electronics, and Information Engineering, Nagaoka University of Technology, Nagaoka, Japan.
Research Organization for Regional Alliance, Kochi University of Technology, Kochi, Japan.
Front Aging Neurosci. 2025 May 27;17:1462951. doi: 10.3389/fnagi.2025.1462951. eCollection 2025.
Identifying older drivers at risk of critical decline in driving safety performance (DSP) is essential for traffic safety. Regional cerebral gray matter (GM) volume may serve as a biomarker for such decline, but its predictive value in real-world driving contexts remains unclear.
We enrolled 94 cognitively healthy older drivers (45 males, 49 females; mean age 77.66 ± 3.67 years) who completed a standardized driving assessment using actual vehicles on a closed-circuit course. DSP was evaluated across six categories: visual search behavior, speeding, indicator signaling, vehicle stability, positioning, and steering. Scores were assigned by a certified driving instructor, with lower scores (<15th percentile) indicating critical DSP decline. Regional GM volumes were quantified using voxel-based morphometry of MRI scans. Feature selection and classification were performed using the Random Forest machine learning algorithm, optimized to identify the most predictive GM regions.
Out of 114 GM regions, eleven were selected as optimal predictors: left angular gyrus, frontal operculum, occipital fusiform gyrus, parietal operculum, postcentral gyrus, planum polare, superior temporal gyrus, and right hippocampus, orbital part of the inferior frontal gyrus, posterior cingulate gyrus, and posterior orbital gyrus. These regions are implicated in attention, spatial cognition, visual processing, and somatosensory integration-functions critical for safe driving. The Random Forest model demonstrated high accuracy and specificity, but moderate precision and recall, limiting immediate real-world application.
While regional GM volume shows promise for identifying older drivers at risk of critical DSP decline, predictive performance remains suboptimal for practical implementation. Additional factors, such as neuronal connectivity assessed by functional MRI, may improve predictive accuracy. Nonetheless, MRI-based assessment of brain structure can enhance our understanding of the neural mechanisms underlying driving safety and inform strategies to prevent traffic accidents among older adults.
识别驾驶安全性能（DSP）可能出现严重下降风险的老年驾驶员对于交通安全至关重要。大脑区域灰质（GM）体积可能是这种下降的生物标志物，但其在实际驾驶情境中的预测价值仍不明确。
我们招募了94名认知健康的老年驾驶员（45名男性，49名女性；平均年龄77.66±3.67岁），他们在封闭赛道上使用实际车辆完成了标准化驾驶评估。DSP通过六个类别进行评估：视觉搜索行为、超速驾驶、指示灯信号、车辆稳定性、定位和转向。分数由一名认证驾驶教练给出，分数较低（低于第15百分位数）表明DSP严重下降。使用MRI扫描的基于体素的形态计量学对大脑区域GM体积进行量化。使用随机森林机器学习算法进行特征选择和分类，该算法经过优化以识别最具预测性的GM区域。
在114个GM区域中，有11个被选为最佳预测指标：左侧角回、额盖、枕颞梭状回、顶盖、中央后回、极平面、颞上回、右侧海马体、额下回眶部、后扣带回和眶后回。这些区域与注意力、空间认知、视觉处理和体感整合有关，这些功能对于安全驾驶至关重要。随机森林模型显示出较高的准确性和特异性，但精度和召回率中等，限制了其在实际中的直接应用。
虽然大脑区域GM体积有望识别DSP严重下降风险的老年驾驶员，但预测性能在实际应用中仍不理想。其他因素，如通过功能MRI评估的神经元连接性，可能会提高预测准确性。尽管如此，基于MRI的脑结构评估可以增强我们对驾驶安全潜在神经机制的理解，并为预防老年人交通事故的策略提供信息。
