Scholarlink

Oral Squamous Cell Carcinoma (OSCC), is a prominent category of cancer and often detected at advanced stages, leading to poor survival rates. Advanced image classification techniques aid in early and accurate detection of OSCC, improving diagnosis and treatment outcomes. This work is based on diagnosing OSCC at an early stage. We combine local binary patterns as well as features extracted using convolutional neural networks (CNN) from the dataset. The combined resultant features are then used to train three different classification models using the resulting feature vectors. The classifiers used are support vector machine (SVM), multi-layer perceptron (MLP) and random forest. The study comprises of experimenting on a dataset of two classes, Normal and OSCC images consisting of 89 and 439 images, respectively. The data imbalance is handled through data augmentation. The result demonstrates a test accuracy of 94.89% by the SVM model using only 528 images of 100× magnification images.

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

Supervised Learning Approaches for OSCC Detection: A Comparative Study of SVM, MLP, and Random Forest Classifiers on Histopathological Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rupesh Mandal,
  • Tenzin Thabkhe,
  • Musaddique Anwar,
  • Nupur Choudhury,
  • Anuran Patgiri,
  • Mrinmoy Mayur Choudhury,
  • Muktanjalee Deka,
  • Jyoti Kumar Barman,
  • F. H. A. Shibly

摘要

Oral Squamous Cell Carcinoma (OSCC), is a prominent category of cancer and often detected at advanced stages, leading to poor survival rates. Advanced image classification techniques aid in early and accurate detection of OSCC, improving diagnosis and treatment outcomes. This work is based on diagnosing OSCC at an early stage. We combine local binary patterns as well as features extracted using convolutional neural networks (CNN) from the dataset. The combined resultant features are then used to train three different classification models using the resulting feature vectors. The classifiers used are support vector machine (SVM), multi-layer perceptron (MLP) and random forest. The study comprises of experimenting on a dataset of two classes, Normal and OSCC images consisting of 89 and 439 images, respectively. The data imbalance is handled through data augmentation. The result demonstrates a test accuracy of 94.89% by the SVM model using only 528 images of 100× magnification images.