Scholarlink

<p>Gastrointestinal (GI) disorders constitute a major global healthcare challenge, necessitating meticulous diagnostic methodologies. However, current research faces significant computational challenges that limit the effectiveness of diagnostic tools. Here, we propose an advanced diagnostic pipeline that addresses these limitations by refining diagnostic accuracy, reducing computational costs, and optimising the analysis of colon disorder imagery. Our method begins by extracting meaningful features from colon images using fused deep residual feature maps obtained by combining pretrained MobileNet, ResNet50V2, and a custom convolutional neural network enhanced with spatial attention. These fused features are subsequently refined using an altruism-based Genetic Algorithm (GA) integrated with Late Acceptance Hill Climbing (LAHC), effectively reducing the feature space by approximately 80%. Evaluations conducted on a Wireless Capsule Endoscopy (WCE) dataset for colon cancer demonstrate superior performance, achieving a test classification accuracy of 98.33%, which significantly surpasses existing algorithms. Furthermore, our method exhibits remarkable computational efficiency in feature selection. These results highlight the potential for enhanced clinical decision-making and improved patient outcomes in colon cancer diagnosis. The source code for this work is publicly available at <a href="https://github.com/TheDankkoder/Colon-Disorder-Classification-using-Altruistic-Genetic-Algorithm-based-Fused-Deep-Feature-Selection">https://github.com/TheDankkoder/Colon-Disorder-Classification-using-Altruistic-Genetic-Algorithm-based-Fused-Deep-Feature-Selection</a>.</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

Colon disorder classification using altruistic genetic algorithm based fused deep feature selection method

  • 期刊论文

  • 发表日期：

    2025年12月27日

查看全文
  • Anurup Naskar,
  • Shivam Singh,
  • Ram Sarkar

摘要

Gastrointestinal (GI) disorders constitute a major global healthcare challenge, necessitating meticulous diagnostic methodologies. However, current research faces significant computational challenges that limit the effectiveness of diagnostic tools. Here, we propose an advanced diagnostic pipeline that addresses these limitations by refining diagnostic accuracy, reducing computational costs, and optimising the analysis of colon disorder imagery. Our method begins by extracting meaningful features from colon images using fused deep residual feature maps obtained by combining pretrained MobileNet, ResNet50V2, and a custom convolutional neural network enhanced with spatial attention. These fused features are subsequently refined using an altruism-based Genetic Algorithm (GA) integrated with Late Acceptance Hill Climbing (LAHC), effectively reducing the feature space by approximately 80%. Evaluations conducted on a Wireless Capsule Endoscopy (WCE) dataset for colon cancer demonstrate superior performance, achieving a test classification accuracy of 98.33%, which significantly surpasses existing algorithms. Furthermore, our method exhibits remarkable computational efficiency in feature selection. These results highlight the potential for enhanced clinical decision-making and improved patient outcomes in colon cancer diagnosis. The source code for this work is publicly available at https://github.com/TheDankkoder/Colon-Disorder-Classification-using-Altruistic-Genetic-Algorithm-based-Fused-Deep-Feature-Selection.