Scholarlink

<p>Lung and Colon Cancers (LCC) are primary reasons of impermanence though imaging investigative techniques are at an advanced stage, detailed cellular-level examinations continue to require gold-standard histopathology. Traditional methods are often time-consuming and costly, leading to delays in diagnosis and potential inter-observer errors. To tackle these challenges, this research proposes a novel Multi-head Attention-based Conditional progressive Generative Adversarial Network (MACGAN) approach to analyze histopathological images effectively. This approach integrates wavelet denoising for preprocessing to effectively remove noise while preserving crucial information, enhancing image quality before classification. Then, the introduced model classifies lung colon cancer using a combination of the multi-head attention mechanism and conditional progressive generative adversarial network, which is specifically combined to strengthen the crucial details of the histopathology images and improve the classification accuracy. Subsequently, the Adaptive Trees Social Relations Optimization (ATSRO) algorithm was utilized to significantly enhance computational efficiency and model performance. Moreover, an accessible graphical user interface has been created to show the outcome of the model. The experimentations are directed through Python on Lung and Colon cancer histopathological image (LC25000) dataset, demonstrate robust performance, achieving 99.95% accuracy,99.95% precision,99.94% sensitivity, 99.94% F-measure, 99.98% specificity and 99.93% Matthews correlation coefficient. Additionally, the proposed method achieves superior efficiency, with the lowest computation time of 30&#xa0;s and a minimal number of trainable parameters (329,863), surpassing previous techniques. Additionally, color space transformations are evaluated across three different datasets to assess their impact on model performance. These advantages underscore the method's suitability for histopathology image analysis, where computational efficiency and parameter optimization are critical. Accordingly, this research offers a significant advancement in histopathological image analysis, providing a highly efficient and accurate tool for early cancer exposure.</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

Multi head attention based conditional progressive GAN for colon cancer histopathological images analysis

  • 期刊论文

  • 发表日期：

    2025年3月21日

查看全文
  • Harikrishna Mulam,
  • Venkata Rambabu Chikati,
  • Anita Kulkarni

摘要

Lung and Colon Cancers (LCC) are primary reasons of impermanence though imaging investigative techniques are at an advanced stage, detailed cellular-level examinations continue to require gold-standard histopathology. Traditional methods are often time-consuming and costly, leading to delays in diagnosis and potential inter-observer errors. To tackle these challenges, this research proposes a novel Multi-head Attention-based Conditional progressive Generative Adversarial Network (MACGAN) approach to analyze histopathological images effectively. This approach integrates wavelet denoising for preprocessing to effectively remove noise while preserving crucial information, enhancing image quality before classification. Then, the introduced model classifies lung colon cancer using a combination of the multi-head attention mechanism and conditional progressive generative adversarial network, which is specifically combined to strengthen the crucial details of the histopathology images and improve the classification accuracy. Subsequently, the Adaptive Trees Social Relations Optimization (ATSRO) algorithm was utilized to significantly enhance computational efficiency and model performance. Moreover, an accessible graphical user interface has been created to show the outcome of the model. The experimentations are directed through Python on Lung and Colon cancer histopathological image (LC25000) dataset, demonstrate robust performance, achieving 99.95% accuracy,99.95% precision,99.94% sensitivity, 99.94% F-measure, 99.98% specificity and 99.93% Matthews correlation coefficient. Additionally, the proposed method achieves superior efficiency, with the lowest computation time of 30 s and a minimal number of trainable parameters (329,863), surpassing previous techniques. Additionally, color space transformations are evaluated across three different datasets to assess their impact on model performance. These advantages underscore the method's suitability for histopathology image analysis, where computational efficiency and parameter optimization are critical. Accordingly, this research offers a significant advancement in histopathological image analysis, providing a highly efficient and accurate tool for early cancer exposure.