基于人工智能算法预测针刺治疗脑卒中患者上肢痉挛的预后研究

Upper extremity spasticity seriously affects the working and living ability of stroke patients. Acupuncture therapy, a kind of traditional Chinese medicine, has a wide range of applications in the antispasmodic. However, a small number of patients suffered increased spasticity after receiving acupuncture. How to identify such patients and optimize rehabilitation programs have become the key issues in clinical decision-making. Our previous study suggested a close relationship between acupuncture treatment effects and central plasticity. However, whether this relationship can be used to predict the prognosis of patients and guide clinical practice still needs further study. To this end, we propose a hypothesis: the use of artificial intelligence algorithms and neuroimaging data could predict the prognosis of acupuncture antispasmodic, and optimize the rehabilitation strategy. The project utilizes functional magnetic resonance imaging to extract central plasticity characteristics, along with artificial intelligence algorithm to predict the prognosis of patients with acupuncture treatment. With the help of data-driven model, we would acquire prognosis before treatment for a possibility of optimizing the rehabilitation strategy. The scientific significance of this project lies in the innovative use of artificial intelligence algorithms to analyze the data and extract the central plasticity characteristics of individuals, with which we could predict the individual-level prognosis. This data-driven prediction does not involve the neural mechanism of spasticity or acupuncture. Its feasibility is outstanding and has good clinical promotion prospects that can be used by rehabilitation physicians to optimize the treatment of acupuncture antispasmodic.

上肢痉挛严重影响脑卒中患者工作和生活能力。传统中医的针刺疗法在解痉方面有广泛的应用，但有一小部分患者接受针刺后痉挛程度反而加重。如何甄别出这类患者，优化康复方案成为临床决策中的关键问题。我们的前期研究提示了针刺治疗效应与中枢可塑性之间的密切联系。但能否运用这种联系来预测患者预后并指导临床实践有待深入研究。为此，我们提出假说：利用人工智能算法和神经影像数据能够预测针刺解痉的预后，实现康复策略的优化。本项目运用功能磁共振检测患者中枢可塑性特征，配合人工智能算法对患者针刺治疗的预后进行预测，在数据驱动模型的帮助下实现治疗前对预后前景的把握，从而优化康复策略。本项目的科学意义在于，创新性地运用人工智能算法分析数据，提取个体的中枢可塑性特征，实现个体水平的预后预测。这种数据驱动的预测过程不涉及痉挛或针刺的神经机制，有良好的可行性和临床推广前景，可助康复医师优化针刺解痉的治疗方案。

目前，卒中后上肢痉挛的发病机制不明。根据既往研究，卒中后痉挛的发生可能与中枢的不良重塑密切相关。本项目运用传统中医的针刺疗法在解痉方面的先验成果，结合中枢功能影像学和数据模型，对患者的解痉预后进行数据驱动的预测。本项目发现，基底节网络作为针刺治疗效应的重要中枢区域，以该子网络的各fMRI参数作为特征矩阵进行预后判断有较强的可行性，这不但印证了针刺解痉与基底节网络的可塑性调整有着强关联性，还定量地将该特征的变化与患者预后建立了关联。相比假干预的对照组，我们发现针刺降低了基底节子网络中的信息传导效率，并且在动态因果模型中对关键节点-下额叶区域起到了抑制作用，通过降低下额叶区-整体降低了网络内神经活动的兴奋性，压制了不良重塑带来的额外神经传导。对该结果的解读有助于完善针刺解除痉挛的中枢机制。.在应用方面，本项目进一步运用上述结论，建立以基底节网络为主要组成的特征矩阵，分别探索了不同特征组成模式，不同数据模型对预后判断的正确性的影响。结果发现，以体素水平的原始信号为特征，采用支持向量机和卷积神经网络模型能达到较好的判别效果。本项目的成功实施，有助于优化后期临床推广时采用的模型及特征空间，有效提升了数据模型推断的效能，为针刺解痉在康复治疗中的应用做出了贡献。