Scholarlink

<p>Cervical cancer is a significant contributor to mortality among women around the globe. But diagnosing it early can be pivotal in saving female’s lives. Nowadays, alongside medical screening, medical professionals are utilizing computer-aided technologies such as machine and deep learning for cervical cancer detection. This study focuses on proposing an automated framework for cervical cancer diagnosis through the integration of hybrid deep feature extraction and ensemble machine learning techniques. Utilizing a comprehensive dataset of Pap smear images, feature extraction was conducted using various pre-trained convolutional neural network (CNN) models such as ResNet101, ResNet50, DenseNet169, DenseNet201, DenseNet121, VGG16, VGG19, and MobileNet, as well as through their hybrid combinations. Further dimensionality reduction via principal component analysis (PCA) was applied to preserve 95% of the variance, thereby reducing computational complexity while maintaining critical information. Afterward, the performance of multiple machine learning classifiers, including SVM, KNN, MLP, LR, RF, XGB, and their ensemble versions (hard voting and soft voting) using high pass filtering was evaluated through these extracted features. Our proposed model, which combines ResNet101, DenseNet121, and DenseNet169 with an ensemble voting classifier (soft), reached a classification accuracy of 97.03% for two classes and 96.67% for five classes of cervical cancer. The results show substantial enhancements in diagnostic accuracy, highlighting the efficacy of hybrid feature extraction and ensemble learning methods for cervical cancer diagnosis.</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

Cervical cancer detection through Pap smear images using hybrid deep feature extraction and ensemble machine learning

  • 期刊论文

  • 发表日期：

    2025年3月22日

查看全文
  • Rajwinder Singh,
  • Hardeep Kaur,
  • Jyoteesh Malhotra

摘要

Cervical cancer is a significant contributor to mortality among women around the globe. But diagnosing it early can be pivotal in saving female’s lives. Nowadays, alongside medical screening, medical professionals are utilizing computer-aided technologies such as machine and deep learning for cervical cancer detection. This study focuses on proposing an automated framework for cervical cancer diagnosis through the integration of hybrid deep feature extraction and ensemble machine learning techniques. Utilizing a comprehensive dataset of Pap smear images, feature extraction was conducted using various pre-trained convolutional neural network (CNN) models such as ResNet101, ResNet50, DenseNet169, DenseNet201, DenseNet121, VGG16, VGG19, and MobileNet, as well as through their hybrid combinations. Further dimensionality reduction via principal component analysis (PCA) was applied to preserve 95% of the variance, thereby reducing computational complexity while maintaining critical information. Afterward, the performance of multiple machine learning classifiers, including SVM, KNN, MLP, LR, RF, XGB, and their ensemble versions (hard voting and soft voting) using high pass filtering was evaluated through these extracted features. Our proposed model, which combines ResNet101, DenseNet121, and DenseNet169 with an ensemble voting classifier (soft), reached a classification accuracy of 97.03% for two classes and 96.67% for five classes of cervical cancer. The results show substantial enhancements in diagnostic accuracy, highlighting the efficacy of hybrid feature extraction and ensemble learning methods for cervical cancer diagnosis.