基于时序卷积网络的早期帕金森多模态检测系统

帕金森病是最常见的神经退行性疾病之一,其临床特征与其他神经退行性疾病有重叠,且缺乏明确的病理机制,导致早期诊断检测困难、误诊率高等问题;为了研究有效的早期帕金森病检测方法,深入探索帕金森病发展的时间特征规律,并提高早期帕金森病预测、分析和诊断决策的准确性,设计了一种基于时序卷积网络的早期帕金森病多模态检测系统,为及时发现早期帕金森病提供辅助诊断依据;该系统利用语音、步态和受试者自测数据,采用多元线性池化方法进行多模态融合,结合时间卷积网络和参数共享方式,以提高系统的检测精度并降低过拟合风险;实验测试结果显示,基于时序卷积网络的早期帕金森病检测系统的准确率达到96.22%,在多项评估指标上优于传统的帕金森检测模型,展现出良好的早期帕金森联合检测效果。

Early Parkinson"s Multimodal Detection System Based on Temporal Convolutional Networks

Parkinson''s disease is one of the most common neurodegenerative diseases, with clinical features overlapping with other neurodegenerative diseases and a lack of precise pathological mechanisms, leading to difficulties in early diagnosis and high misdiagnosis rates.; In order to study effective early detection methods for Parkinson''s disease, deeply explore the temporal characteristics of Parkinson''s disease development, and improve the accuracy of early Parkinson''s disease prediction, analysis, and diagnostic decision-making, a multimodal detection system for early Parkinson''s disease based on temporal convolutional networks was designed, providing an auxiliary diagnostic basis for the timely detection of early Parkinson''s disease; The system utilizes speech, gait, and subject self-test data, adopts multiple linear pooling methods for multimodal fusion, and combines time convolutional networks and parameter sharing methods to improve the detection accuracy of the system and reduce overfitting risks; The results of ablation and comparative experiments show that the accuracy of the early Parkinson''s disease detection system based on temporal convolutional networks reaches 96.22%, which is superior to traditional Parkinson''s detection models in multiple evaluation indicators and demonstrates good early Parkinson''s joint detection performance.