Scholarlink

<p>This paper presents a completely computer-assisted automated system for detecting and diagnosing cervical cancer, utilizing advanced deep learning Convolutional Neural Network (CNN) technology. The proposed cervical cancer detection system employs two distinct CNN architectures for classifying cervical images. The first CNN architecture extracts extrinsic features from augmented datasets of both healthy and cancerous cervical images. These extracted features and measurements are then input into the second CNN to execute the classification procedure. The second CNN generates classification outputs that categorize images as either healthy or cancerous. Subsequently, Erosion-Dilation (ED) mathematical operations are applied to images identified as cancerous to identify cancer pixels. The cancerous pixel regions are analyzed to characterize the abnormal areas present in the cervigram images. These segmented regions are utilized to study the structural and textural properties of cancer-affected regions. The proposed Dual Order Convolutional Neural Networks (DOCNN) based cervical cancer detection system achieves performance metrics of 98.92% Sensitivity, 98.85% Specificity, 98.93% Positive Predictive Value (PPV), 98.81% Negative Predictive Value (NPV), 1.15% False Positive Rate (FPR), 1.08% False Negative Rate (FNR), 71.03 Positive Likelihood Ratio (PLR), and 0.04167 Negative Likelihood Ratio (NLR) when tested on cervical images from the Kaggle dataset.</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 and Diagnosis System using Dual Order Convolutional Neural Networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月11日

查看全文
  • M. R. Christhu Raj,
  • S. Baskar,
  • C. Kumar,
  • Abhinandan Routray

摘要

This paper presents a completely computer-assisted automated system for detecting and diagnosing cervical cancer, utilizing advanced deep learning Convolutional Neural Network (CNN) technology. The proposed cervical cancer detection system employs two distinct CNN architectures for classifying cervical images. The first CNN architecture extracts extrinsic features from augmented datasets of both healthy and cancerous cervical images. These extracted features and measurements are then input into the second CNN to execute the classification procedure. The second CNN generates classification outputs that categorize images as either healthy or cancerous. Subsequently, Erosion-Dilation (ED) mathematical operations are applied to images identified as cancerous to identify cancer pixels. The cancerous pixel regions are analyzed to characterize the abnormal areas present in the cervigram images. These segmented regions are utilized to study the structural and textural properties of cancer-affected regions. The proposed Dual Order Convolutional Neural Networks (DOCNN) based cervical cancer detection system achieves performance metrics of 98.92% Sensitivity, 98.85% Specificity, 98.93% Positive Predictive Value (PPV), 98.81% Negative Predictive Value (NPV), 1.15% False Positive Rate (FPR), 1.08% False Negative Rate (FNR), 71.03 Positive Likelihood Ratio (PLR), and 0.04167 Negative Likelihood Ratio (NLR) when tested on cervical images from the Kaggle dataset.