Scholarlink

<p>Skin cancer remains one of the most prevalent and life-threatening conditions worldwide, emphasizing the critical need for accurate and efficient diagnostic systems. This research introduces a hybrid classification model integrating Multi-Objective Genetic Algorithms (MOGA) and Artificial Neural Networks (ANN) to address the challenges of noisy datasets, feature complexity, and the trade-off between accuracy and model simplicity. Leveraging Pareto optimization, the MOGA-ANN framework dynamically optimizes the neural network connection weights while keeping the architecture fixed, enhancing its ability to detect intricate patterns in medical imagery. The proposed framework was rigorously evaluated using the Skin Cancer ISIC dataset, achieving an impressive testing accuracy of 97.3%. This result highlights the model’s capability to generalize across diverse scenarios and improve diagnostic precision. We evaluate our approach on public dermoscopic datasets (ISIC2018 and HAM10000) and treat the training as a bi-objective optimization: minimizing validation error while minimizing the fraction of non-zero weights in the ANN head. The optimization uses a simplified multi-objective genetic algorithm inspired by NSGA-II to produce a Pareto set of models that trade off accuracy and sparsity. By combining the adaptability of genetic algorithms with the learning power of neural networks, this study offers a novel and effective solution for automated medical diagnostics, particularly in skin cancer detection.</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

Towards Accurate Skin Cancer Diagnosis: A Hybrid MOGA-ANN Model for Precision Classification and Detection

  • 期刊论文

  • 发表日期：

    2026年4月13日

查看全文
  • Sanket Dan,
  • Jayshree Bhattacharya,
  • Raja Dey,
  • Srinjoy Acharjee,
  • Saptarshi Chakraborty,
  • Shrishti Bit,
  • Tandrima Biswas

摘要

Skin cancer remains one of the most prevalent and life-threatening conditions worldwide, emphasizing the critical need for accurate and efficient diagnostic systems. This research introduces a hybrid classification model integrating Multi-Objective Genetic Algorithms (MOGA) and Artificial Neural Networks (ANN) to address the challenges of noisy datasets, feature complexity, and the trade-off between accuracy and model simplicity. Leveraging Pareto optimization, the MOGA-ANN framework dynamically optimizes the neural network connection weights while keeping the architecture fixed, enhancing its ability to detect intricate patterns in medical imagery. The proposed framework was rigorously evaluated using the Skin Cancer ISIC dataset, achieving an impressive testing accuracy of 97.3%. This result highlights the model’s capability to generalize across diverse scenarios and improve diagnostic precision. We evaluate our approach on public dermoscopic datasets (ISIC2018 and HAM10000) and treat the training as a bi-objective optimization: minimizing validation error while minimizing the fraction of non-zero weights in the ANN head. The optimization uses a simplified multi-objective genetic algorithm inspired by NSGA-II to produce a Pareto set of models that trade off accuracy and sparsity. By combining the adaptability of genetic algorithms with the learning power of neural networks, this study offers a novel and effective solution for automated medical diagnostics, particularly in skin cancer detection.