Scholarlink

Cervical cancer ranks as the fourth-most common cancer among women. It is caused by the infection of the human papillomavirus (HPV), and this virus is present in over 99% of cervical cancers. Prompt diagnosis and treatment can significantly increase a patient’s survival rate. Hence, it is crucial to screen for cervical cancer on a regular basis. This paper investigates the potential of a deep learning model for differentiating between healthy and cancerous cells. The automatic classification of cervical images can be helpful for quick diagnosis and reducing work overload. This study implements and conducts a comparative analysis of several deep neural network architectures which includes VGGNet (VGG16), ResNet (Resnet-50), EfficientNet (efficientnet-B0, efficientnet-B1, efficientnet-B3), DenseNet169 and MobileNetV2 trained on publicly available Mendeley LBC dataset. Out of the various models, it is found that MobileNetV2 combined with preprocessing techniques, achieved a test accuracy of 100.00% and it stands out as the best performing model.

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

Automatic Classification of Cervical Dysplasia Using Deep Neural Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jheelam Mondal,
  • Rajdeep Chatterjee,
  • Mahendra Kumar Gourisaria

摘要

Cervical cancer ranks as the fourth-most common cancer among women. It is caused by the infection of the human papillomavirus (HPV), and this virus is present in over 99% of cervical cancers. Prompt diagnosis and treatment can significantly increase a patient’s survival rate. Hence, it is crucial to screen for cervical cancer on a regular basis. This paper investigates the potential of a deep learning model for differentiating between healthy and cancerous cells. The automatic classification of cervical images can be helpful for quick diagnosis and reducing work overload. This study implements and conducts a comparative analysis of several deep neural network architectures which includes VGGNet (VGG16), ResNet (Resnet-50), EfficientNet (efficientnet-B0, efficientnet-B1, efficientnet-B3), DenseNet169 and MobileNetV2 trained on publicly available Mendeley LBC dataset. Out of the various models, it is found that MobileNetV2 combined with preprocessing techniques, achieved a test accuracy of 100.00% and it stands out as the best performing model.