Scholarlink

Purpose <p>This study proposes a method for detecting anomalies in health data using Markovian geometric diffusion, aiming to enhance clinical diagnoses. It addresses the challenge of identifying outlier patterns that significantly deviate from the majority in various medical datasets.</p> Methods <p>The proposed approach quantifies similarity between instances through probabilities derived from geometric diffusion, enabling robust identification of atypical values. Its performance was assessed on seven real-world medical datasets, including arrhythmia, mammography, cardiotocography, and lymphography, retrieved from public repositories.</p> Results <p>Experimental findings revealed superior accuracy and fewer false positives compared to traditional outlier detection methods, particularly in scenarios characterized by high dimensionality and imbalanced classes. DA-DGM consistently identified unusual patterns, demonstrating a reliable balance between detection sensitivity and specificity across the evaluated datasets.</p> Conclusion <p>In conclusion, the Markovian geometric diffusion–based technique proves a promising tool for aiding decision-making in healthcare, potentially enhancing early detection of pathologies and improving patient outcomes. Future research should explore parameter optimization, scalability, and broader clinical applications to further validate its efficacy.</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

Anomaly detection in healthcare data based on Markovian geometric diffusion: an advanced approach for medical diagnostics

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • E. C. Ramos,
  • L. C. de Souza,
  • G. H. M. B. Motta

摘要

Purpose

This study proposes a method for detecting anomalies in health data using Markovian geometric diffusion, aiming to enhance clinical diagnoses. It addresses the challenge of identifying outlier patterns that significantly deviate from the majority in various medical datasets.

Methods

The proposed approach quantifies similarity between instances through probabilities derived from geometric diffusion, enabling robust identification of atypical values. Its performance was assessed on seven real-world medical datasets, including arrhythmia, mammography, cardiotocography, and lymphography, retrieved from public repositories.

Results

Experimental findings revealed superior accuracy and fewer false positives compared to traditional outlier detection methods, particularly in scenarios characterized by high dimensionality and imbalanced classes. DA-DGM consistently identified unusual patterns, demonstrating a reliable balance between detection sensitivity and specificity across the evaluated datasets.

Conclusion

In conclusion, the Markovian geometric diffusion–based technique proves a promising tool for aiding decision-making in healthcare, potentially enhancing early detection of pathologies and improving patient outcomes. Future research should explore parameter optimization, scalability, and broader clinical applications to further validate its efficacy.