Scholarlink

The COVID-19 epidemic has disproportionately affected sufferer with COVID-19, pneumonia, and other respiratory conditions, increasing their risk of severe illness and mortality while disrupting essential care. The overlap in symptoms and imaging findings among these diseases presents diagnostic challenges, potentially leading to misdiagnosis and delayed treatment. This study introduces a novel approach using deep learning, specifically deep autoencoders, to denoise COVID-19, normal, and pneumonia images. This paper compares three autoencoder models—Convolutional Denoising Autoencoders (CDAE), Stacked Denoising Autoencoders (SDAE), and a Novel Symlet-based Denoising Autoencoder (NSDAE). With a high Structural Similarity Index (SSIM) of 0.7885 and performance validation accuracies of 0.9602 and 0.9650 through the use of RELU implementation with the ADAM optimizer, the NSDAE model showed exceptional performance on important quality measures. These results demonstrate that the NSDAE model may enhance pneumonia and COVID-19 decision-making and normalization, leading to more precise diagnosis and better in-patient outcomes.

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

Novel Symlet Denoising Autoencoder for CT Diagnosis of Respiratory Diseases

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • S. Nivetha,
  • H. Hannah Inbarani

摘要

The COVID-19 epidemic has disproportionately affected sufferer with COVID-19, pneumonia, and other respiratory conditions, increasing their risk of severe illness and mortality while disrupting essential care. The overlap in symptoms and imaging findings among these diseases presents diagnostic challenges, potentially leading to misdiagnosis and delayed treatment. This study introduces a novel approach using deep learning, specifically deep autoencoders, to denoise COVID-19, normal, and pneumonia images. This paper compares three autoencoder models—Convolutional Denoising Autoencoders (CDAE), Stacked Denoising Autoencoders (SDAE), and a Novel Symlet-based Denoising Autoencoder (NSDAE). With a high Structural Similarity Index (SSIM) of 0.7885 and performance validation accuracies of 0.9602 and 0.9650 through the use of RELU implementation with the ADAM optimizer, the NSDAE model showed exceptional performance on important quality measures. These results demonstrate that the NSDAE model may enhance pneumonia and COVID-19 decision-making and normalization, leading to more precise diagnosis and better in-patient outcomes.