By 2050, a quarter of the US population will be over the age of 65 with greater than a 40% risk of developing life-altering neuromusculoskeletal pathologies. The potential of wearables, such as Apple AirPods and hearing aids, to provide personalized preventative and predictive health monitoring outside of the clinic is nascent, but large quantities of open-ended data that capture movement in the physical world now exist. Algorithms that leverage existing wearable technology to detect subtle changes to walking mechanics, an early indicator of neuromusculoskeletal pathology, have successfully been developed to determine population-level statistics, but individual-level variability is more difficult to parse from population-level data. Like genetic sequencing, the individual's gait pattern can be discerned by decomposing the movement signal into its fundamental features from which we can detect "mutations" or changes to the pattern that are early indicators of pathology - movement-based biomarkers. We have developed a novel approach to quantify "normal baseline movement" at an individual level, combining methods from gait laboratories with methods used to characterize stellar oscillations. We tested our approach by asking participants to complete an outdoor circuit while wearing a pair of AirPods, using orthopaedic braces to simulate pathology. We found that the novel features we propose are sensitive enough to distinguish between normal walking and brace walking at the population level and at the individual level in all sensor directions (both p $<$ 0.05). We also perform principal component analysis on our population-level and individual-level models, and find significant differences between individuals as well as between the overall population model and most individuals. We also demonstrate the potential of these gait features in deep learning applications.

翻译：到2050年，美国四分之一人口将超过65岁，其中超过40%面临罹患改变生命的神经肌肉骨骼疾病的风险。以Apple AirPods和助听器为代表的可穿戴设备在临床环境外提供个性化预防和预测性健康监测的潜力尚处萌芽阶段，但目前已存在大量捕捉现实世界运动行为的开放式数据。利用现有可穿戴技术检测行走力学细微变化的算法已成功开发，可用于确定群体层面的统计数据——这些细微变化正是神经肌肉骨骼疾病的早期指标。然而个体层面的变异更难从群体数据中解析。与基因测序类似，个体的步态模式可通过将运动信号分解为基本特征来识别，从中我们能检测到作为病理早期指标的“突变”或模式变化——即基于运动的生物标志物。我们开发了一种在个体层面量化“正常基线运动”的新方法，该方法融合了步态实验室技术与恒星振荡表征技术。我们通过要求参与者佩戴AirPods完成户外行走测试来验证该方法，并使用骨科支具模拟病理状态。研究发现，我们提出的新特征具有足够灵敏度，能在群体层面和个体层面区分正常行走与支具行走（所有传感器方向p值均小于0.05）。我们对群体模型和个体模型进行主成分分析，发现个体间存在显著差异，且总体群体模型与大多数个体模型也存在显著差异。我们还展示了这些步态特征在深度学习应用中的潜力。