Scholarlink

<p>Effective management of the enormous volume of medical images for storage and transmission is crucial, particularly in telemedicine applications, where medical image compression and encryption play a significant role. In medical imaging, Computed Tomography is the famed tool for diagnosis. Prior, diverse research work has been conducted in the field of telemedicine for compressing and encrypting Computed Tomography images. However, the schemes have less compression ratio, more encryption time, and poor performance. To address these difficulties, this research proposes a novel DeepTeleNet strategy for effectively compressing and encrypting the Computed Tomography image. The DeepTeleNet framework utilizes a Hybrid Dilated Convolution U-Net for the segmentation of Region of Interest-based computed Tomography image that enables the model to capture more spatial information. The Hybrid Dilated convolution U-Net framework incorporates an Adaptive Attention Fusion Module, which mitigates the irrelevant information and enhances the feature representation in the Region of Interest area by the utilization of the Channel Attention Module and Spatial Attention Module. Additionally, the proposed strategy deployed an Adaptive Arithmetic and Adaptive Huffman encoding for Region of Interest lossless compression, where the Adaptive Arithmetic encoding dynamically adjusts the probability estimates as per the observed frequencies and the Adaptive Huffman encoding dynamically updates the trees as data is processed. Moreover, the enhanced Autoencoder is employed for the lossy compression of non- Region of Interest that integrates the Convolutional Autoencoder network and Principal Component Analysis. Further, the proposed strategy utilizes multiple chaotic maps with Sparse Lifting Wavelet Transform to authenticate Computed Tomography images by embedding an encrypted watermark into the images. The experimental result demonstrates that the DeepTeleNet scheme effectively compressed and encrypted the computed Tomography images and achieved a higher compression ratio, and lower encryption time of 57.8241%, and 0.09&#xa0;s compared to existing methodologies, underscores its effectiveness in Computed Tomography image compression and encryption, making it a promising solution for telemedicine application.</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

Deep learning-based compression and encryption of CT images for secure telemedicine applications

  • 期刊论文

  • 发表日期：

    2025年1月29日

查看全文
  • S. Rosaline,
  • D. Paulraj

摘要

Effective management of the enormous volume of medical images for storage and transmission is crucial, particularly in telemedicine applications, where medical image compression and encryption play a significant role. In medical imaging, Computed Tomography is the famed tool for diagnosis. Prior, diverse research work has been conducted in the field of telemedicine for compressing and encrypting Computed Tomography images. However, the schemes have less compression ratio, more encryption time, and poor performance. To address these difficulties, this research proposes a novel DeepTeleNet strategy for effectively compressing and encrypting the Computed Tomography image. The DeepTeleNet framework utilizes a Hybrid Dilated Convolution U-Net for the segmentation of Region of Interest-based computed Tomography image that enables the model to capture more spatial information. The Hybrid Dilated convolution U-Net framework incorporates an Adaptive Attention Fusion Module, which mitigates the irrelevant information and enhances the feature representation in the Region of Interest area by the utilization of the Channel Attention Module and Spatial Attention Module. Additionally, the proposed strategy deployed an Adaptive Arithmetic and Adaptive Huffman encoding for Region of Interest lossless compression, where the Adaptive Arithmetic encoding dynamically adjusts the probability estimates as per the observed frequencies and the Adaptive Huffman encoding dynamically updates the trees as data is processed. Moreover, the enhanced Autoencoder is employed for the lossy compression of non- Region of Interest that integrates the Convolutional Autoencoder network and Principal Component Analysis. Further, the proposed strategy utilizes multiple chaotic maps with Sparse Lifting Wavelet Transform to authenticate Computed Tomography images by embedding an encrypted watermark into the images. The experimental result demonstrates that the DeepTeleNet scheme effectively compressed and encrypted the computed Tomography images and achieved a higher compression ratio, and lower encryption time of 57.8241%, and 0.09 s compared to existing methodologies, underscores its effectiveness in Computed Tomography image compression and encryption, making it a promising solution for telemedicine application.