Scholarlink

<p>Parkinson’s disease (PD) disrupts episodic memory, undermining patients’ well-being. Yet, standard assessments consist in counts of recalled information units, limiting informativeness and scalability. To refine testing, we recruited 74 participants (35 PD patients, 39 controls) who retold an action and a non-action story. We analyzed their verbosity (total and content word production), semantic acuity (conceptual similarity to the stories), and organizational similarity (structural similarity to the stories) using part-of-speech tagging, the cosine similarity of text embeddings, and speech graphs. People with PD were less verbose only when retelling the action story, and they exhibited lower semantic acuity and organizational similarity across stories. Our approach (i) achieved robust discrimination between patients and controls (ROC-AUC = 0.79), mostly driven by action-related metrics; (ii) outperformed traditional cognitive, executive, and episodic memory tests scores (ROC-AUC = 0.70); and (iii) distinguished between PD-MCI and PD-nMCI. Episodic anomalies thus manifest across multiple dimensions that elude standard instruments but can be tapped digitally to enhance cognitive assessments in PD.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Digitizing episodic memory assessments in Parkinson’s disease via natural language processing

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月22日

查看全文
  • Lucas Federico Sterpin,
  • Jeremías Inchauspe,
  • Camilo Avendaño Avello,
  • Franco J. Ferrante,
  • Gonzalo Nicolás Pérez,
  • Edinson Muñoz,
  • Pedro Chaná-Cuevas,
  • Yamile Bocanegra,
  • Leonardo Moreno,
  • Lucia Amoruso,
  • Adolfo M. García

摘要

Parkinson’s disease (PD) disrupts episodic memory, undermining patients’ well-being. Yet, standard assessments consist in counts of recalled information units, limiting informativeness and scalability. To refine testing, we recruited 74 participants (35 PD patients, 39 controls) who retold an action and a non-action story. We analyzed their verbosity (total and content word production), semantic acuity (conceptual similarity to the stories), and organizational similarity (structural similarity to the stories) using part-of-speech tagging, the cosine similarity of text embeddings, and speech graphs. People with PD were less verbose only when retelling the action story, and they exhibited lower semantic acuity and organizational similarity across stories. Our approach (i) achieved robust discrimination between patients and controls (ROC-AUC = 0.79), mostly driven by action-related metrics; (ii) outperformed traditional cognitive, executive, and episodic memory tests scores (ROC-AUC = 0.70); and (iii) distinguished between PD-MCI and PD-nMCI. Episodic anomalies thus manifest across multiple dimensions that elude standard instruments but can be tapped digitally to enhance cognitive assessments in PD.