Scholarlink

<p>In the field of medical imaging, the ability to capture and process different types of biomedical images for diagnostic and prognostic purposes has progressed tremendously. Computer vision and artificial intelligence have the potential to exceed human diagnostic abilities by uncovering hidden information in these images. However, current systems often face challenges such as low classification accuracy, high computational costs, and complexity. This research introduces an ensemble machine learning approach for efficient and accurate classification of medical images, including CT kidney scans, chest X-rays, and brain MRIs. Image quality is enhanced in the preprocessing stage using a modified Wiener filter embedded with a finite impulse response filter. Then, segmentation is handled by the SegFormer V-shaped network, which effectively isolates infected regions. Subsequently, local and global attributes of the image, as well as deep features, are extracted. Following that, the Enhanced Crayfish Optimization Algorithm (ECOA) is used for dimensionality reduction to provide an improvement in feature learning. Seven machine learning classifiers are evaluated, with the highest performance observed in the Light Gradient Boosting Machine (LGBM), achieving accuracy rates of 99.32%, 99.70%, and 99.80% on CT kidney, chest X-ray, and brain MRI datasets, respectively. Additionally, SHapley Additive eXplanations (SHAP) is incorporated to improve model transparency, allowing clinicians to better interpret and trust the classification results, crucial for clinical decision-making in medical image analysis.</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

Improving Medical Image Classification with Segmentation-Driven Ensemble Machine Learning

  • 期刊论文

  • 发表日期：

    2025年11月1日

查看全文
  • Krishan Kumar,
  • Sonal Dahiya

摘要

In the field of medical imaging, the ability to capture and process different types of biomedical images for diagnostic and prognostic purposes has progressed tremendously. Computer vision and artificial intelligence have the potential to exceed human diagnostic abilities by uncovering hidden information in these images. However, current systems often face challenges such as low classification accuracy, high computational costs, and complexity. This research introduces an ensemble machine learning approach for efficient and accurate classification of medical images, including CT kidney scans, chest X-rays, and brain MRIs. Image quality is enhanced in the preprocessing stage using a modified Wiener filter embedded with a finite impulse response filter. Then, segmentation is handled by the SegFormer V-shaped network, which effectively isolates infected regions. Subsequently, local and global attributes of the image, as well as deep features, are extracted. Following that, the Enhanced Crayfish Optimization Algorithm (ECOA) is used for dimensionality reduction to provide an improvement in feature learning. Seven machine learning classifiers are evaluated, with the highest performance observed in the Light Gradient Boosting Machine (LGBM), achieving accuracy rates of 99.32%, 99.70%, and 99.80% on CT kidney, chest X-ray, and brain MRI datasets, respectively. Additionally, SHapley Additive eXplanations (SHAP) is incorporated to improve model transparency, allowing clinicians to better interpret and trust the classification results, crucial for clinical decision-making in medical image analysis.