Department of Biomedical Engineering, University of Calgary, Calgary, AB, Canada.
Department of Geomatics Engineering, University of Calgary, Calgary, AB, Canada.
J Alzheimers Dis. 2023;92(4):1487-1497. doi: 10.3233/JAD-221268.
Driving behavior as a digital marker and recent developments in blood-based biomarkers show promise as a widespread solution for the early identification of Alzheimer's disease (AD).
This study used artificial intelligence methods to evaluate the association between naturalistic driving behavior and blood-based biomarkers of AD.
We employed an artificial neural network (ANN) to examine the relationship between everyday driving behavior and plasma biomarker of AD. The primary outcome was plasma Aβ42/Aβ40, where Aβ42/Aβ40 < 0.1013 was used to define amyloid positivity. Two ANN models were trained and tested for predicting the outcome. The first model architecture only includes driving variables as input, whereas the second architecture includes the combination of age, APOE ɛ4 status, and driving variables.
All 142 participants (mean [SD] age 73.9 [5.2] years; 76 [53.5%] men; 80 participants [56.3% ] with amyloid positivity based on plasma Aβ42/Aβ40) were cognitively normal. The six driving features, included in the ANN models, were the number of trips during rush hour, the median and standard deviation of jerk, the number of hard braking incidents and night trips, and the standard deviation of speed. The F1 score of the model with driving variables alone was 0.75 [0.023] for predicting plasma Aβ42/Aβ40. Incorporating age and APOE ɛ4 carrier status improved the diagnostic performance of the model to 0.80 [>0.051].
Blood-based AD biomarkers offer a novel opportunity to establish the efficacy of naturalistic driving as an accessible digital marker for AD pathology in driving research.
驾驶行为作为一种数字标志物，以及基于血液的生物标志物的最新进展，为早期发现阿尔茨海默病（AD）提供了一种广泛的解决方案。
本研究使用人工智能方法评估自然驾驶行为与 AD 基于血液的生物标志物之间的关联。
我们采用人工神经网络（ANN）来研究日常驾驶行为与 AD 血浆生物标志物之间的关系。主要结局是血浆 Aβ42/Aβ40，其中 Aβ42/Aβ40<0.1013 用于定义淀粉样蛋白阳性。训练和测试了两个 ANN 模型来预测结果。第一个模型结构仅将驾驶变量作为输入，而第二个结构将年龄、APOE ε4 状态和驾驶变量结合起来。
所有 142 名参与者（平均[标准差]年龄 73.9[5.2]岁；76[53.5%]名男性；80 名[56.3%]参与者根据血浆 Aβ42/Aβ40 确定为淀粉样蛋白阳性）认知正常。纳入 ANN 模型的 6 个驾驶特征包括高峰时段的出行次数、急动度的中位数和标准差、急刹车事件和夜间出行次数以及速度的标准差。仅使用驾驶变量的模型预测血浆 Aβ42/Aβ40 的 F1 评分为 0.75[0.023]。纳入年龄和 APOE ε4 携带者状态可将模型的诊断性能提高至 0.80[>0.051]。
基于血液的 AD 生物标志物为自然驾驶作为 AD 病理学的一种可及的数字标志物在驾驶研究中的功效提供了一个新的机会。
