Parkinson's disease (PD) is clinically heterogeneous across patients and may be classified in three motor phenotypes: tremor dominant (TD), postural instability and gait disorder (PIGD), and undetermined. Despite the significant clinical characterization of motor phenotypes, little is known about how electrophysiological data, particularly subthalamic nucleus local field potentials (STN‐LFP), differ between TD and PIGD patients. This is relevant since increased STN‐LFP bandpower at α–β range (8–35 Hz) is considered a potential PD biomarker and, therefore, a critical setpoint to drive adaptive deep brain stimulation. Acknowledging STN‐LFP differences between phenotypes, mainly in rest and movement states, would better fit DBS to clinical and motor demands. We studied this issue through spectral analyses on 35 STN‐LFP in TD and PIGD patients during rest and movement. We demonstrated that higher β₂ activity (22–35 Hz) was observed in PIGD only during rest. Additionally, bandpower differences between rest and movement occurred at the α–β range, but with different patterns as per phenotypes: movement‐induced desynchronization concerned lower frequencies in TD (10–20 Hz) and higher frequencies in PIGD patients (21–28 Hz). Finally, when supervised learning algorithms were employed aiming to discriminate PD phenotypes based on STN‐LFP bandpower features, movement information had improved the classification accuracy, achieving peak performances when TD and PIGD movement‐induced desynchronization ranges were considered. These results suggest that STN‐LFP β‐band encodes phenotype‐movement dependent information in PD patients.

中文翻译：

---

帕金森氏病丘脑底核局部场电位的光谱特征：表型和运动物质

帕金森氏病（PD）在患者中临床上异质，可以分为三种运动表型：震颤显性（TD），姿势不稳和步态障碍（PIGD）和不确定。尽管运动表型具有显着的临床特征，但对于TD和PIGD患者之间的电生理数据，尤其是丘脑底核局部场电位（STN-LFP）的差异知之甚少。这很重要，因为在α-β范围（8-35 Hz）上增加的STN-LFP带宽被认为是潜在的PD生物标志物，因此是驱动适应性深部脑刺激的关键设定点。认识到表型之间的STN-LFP差异（主要是在休息和运动状态下）将更好地使DBS适应临床和运动需求。我们通过对TD和PIGD患者在休息和运动期间的35例STN-LFP进行频谱分析研究了此问题。我们证明了更高的β仅在休息期间，在PIGD中观察到_2个活动（22–35 Hz）。此外，静止和运动之间的带功率差异发生在α–β范围，但根据表型具有不同的模式：运动引起的失步涉及TD的较低频率（10–20 Hz）和PIGD患者的较高频率（21–28 Hz） ）。最后，当采用旨在基于STN-LFP功率特征来区分PD表型的监督学习算法时，运动信息提高了分类准确性，在考虑TD和PIGD运动引起的失步范围时达到了峰值性能。这些结果表明，STN-LFPβ带编码PD患者的表型运动依赖性信息。