Scholarlink

<p>This study introduces an enhanced texture-based algorithm for the classification of oral cancer images. The images are first extensively preprocessed to enhance the affected area with techniques like gamma correction, adaptive histogram equalization, and sharpening of images using a Laplacian filter. Then a feature descriptor is extracted using an Adaptive Multiscale Local Mesh ternary patterns, which uses an adaptive threshold, and a sliding window of multiple scales. A machine learning model is then used for classification, which is also optimised using Grey Wolf optimisation. The pedagogy yields an overall average accuracy of 98.29% and 99.89% on two publicly available datasets. Also, the model is compared to three state-of-the-art techniques for oral cancer detection and is found to give an average improvement of 9.21% and 20.67% on the respective datasets.</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

An adaptive multiscale local mesh ternary pattern technique with extensive pre-processing and Grey Wolf optimisation based classifiers for oral cancer image classification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月9日

查看全文
  • Varun Srivastava,
  • Khushi Garg,
  • Samarth Soni,
  • Arun Balodi,
  • Manoj Tolani,
  • Vikash Singh

摘要

This study introduces an enhanced texture-based algorithm for the classification of oral cancer images. The images are first extensively preprocessed to enhance the affected area with techniques like gamma correction, adaptive histogram equalization, and sharpening of images using a Laplacian filter. Then a feature descriptor is extracted using an Adaptive Multiscale Local Mesh ternary patterns, which uses an adaptive threshold, and a sliding window of multiple scales. A machine learning model is then used for classification, which is also optimised using Grey Wolf optimisation. The pedagogy yields an overall average accuracy of 98.29% and 99.89% on two publicly available datasets. Also, the model is compared to three state-of-the-art techniques for oral cancer detection and is found to give an average improvement of 9.21% and 20.67% on the respective datasets.