Scholarlink

<p>Generative models play a pivotal role in speech and audio generation, and recent advancements have expanded their applications in medical diagnosis. This study provides a comprehensive comparative analysis of various generative models, focusing on their impact on the healthcare sector. The primary goal of our research is to emphasize the models’ potential to revolutionize medical diagnostics while also addressing the challenges and risks associated with their implementation. Unlike prior research, which broadly explored generative audio models, this work uniquely highlights their critical importance in the medical domain. The study introduces the fundamental concepts of audio model generation and their applications, followed by a theoretical exploration of recent models that have shown notable success in medical diagnostics. A detailed comparison of model performance is conducted across two distinct medical audio datasets, evaluating key metrics such as accuracy, precision, recall, F1 score, and specificity. For the PIG Disease Dataset, models like MelGAN and SpecGAN demonstrate high accuracy (0.9917) and precision (0.9918), with superior recall and specificity, while WaveGAN also performs remarkably well with an accuracy of 0.992. In the Vascular Access Sounds Dataset, both Differentiable Digital Signal Processing (DDSP) and MelGAN exhibit outstanding performance, with accuracy and precision scores of 0.9917 and specificity of 0.9833. These results showcase the transformative potential of generative models in enhancing diagnostic accuracy across medical applications.</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

Comparative evaluation of generative audio models for precise disease classification via sound-based datasets

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月21日

查看全文
  • Esraa Hassan,
  • Nora El-Rashidy

摘要

Generative models play a pivotal role in speech and audio generation, and recent advancements have expanded their applications in medical diagnosis. This study provides a comprehensive comparative analysis of various generative models, focusing on their impact on the healthcare sector. The primary goal of our research is to emphasize the models’ potential to revolutionize medical diagnostics while also addressing the challenges and risks associated with their implementation. Unlike prior research, which broadly explored generative audio models, this work uniquely highlights their critical importance in the medical domain. The study introduces the fundamental concepts of audio model generation and their applications, followed by a theoretical exploration of recent models that have shown notable success in medical diagnostics. A detailed comparison of model performance is conducted across two distinct medical audio datasets, evaluating key metrics such as accuracy, precision, recall, F1 score, and specificity. For the PIG Disease Dataset, models like MelGAN and SpecGAN demonstrate high accuracy (0.9917) and precision (0.9918), with superior recall and specificity, while WaveGAN also performs remarkably well with an accuracy of 0.992. In the Vascular Access Sounds Dataset, both Differentiable Digital Signal Processing (DDSP) and MelGAN exhibit outstanding performance, with accuracy and precision scores of 0.9917 and specificity of 0.9833. These results showcase the transformative potential of generative models in enhancing diagnostic accuracy across medical applications.