Scholarlink

<p>Cervical cancer (CC) is the most prevalent cancer in women; it is a serious risk to the health of women. The colposcopy procedure is an essential part of CC prevention. However, existing methods often struggle with limited feature representation, poor multimodal fusion, and low accuracy in early-stage CC classification. To overcome these challenges, a novel CERVIC-NET is proposed for accurate classification and staging of CC using multi-modality images. The patient undergoes HPV DNA testing, where a negative result requires no further action, while a positive result leads to clinical screening for detailed cervical examination. The input both colposcopy and Pap smear images are pre-processed using Anisotropic Guided Filter (AGF) is utilized to enhance image clarity and suppress noise. The proposed Dual-Stream Capsule Pyramid Network (DSCP-NET) is employed to extract spatial and hierarchical features from both imaging modalities. The extracted features are fused to capture multimodal correlations and fed into Gated Multi-Layer Perceptron (GMLP) for accurately classify Negative for Squamous Cell Carcinoma (SCC), Low-Grade Squamous Intraepithelial Lesion (LSIL), High-Grade Squamous Intraepithelial Lesion (HSIL), and Intraepithelial Lesion or Malignancy (NILM). The effectiveness of the CERVIC-NET is evaluated utilizing recall, accuracy, precision, specificity, and F1-score. The overall accuracy and precision of the CERVIC-NET are 98.50% and 98.16%, respectively. The CERVIC-NET improves an overall accuracy by 9.62%, 4.86%, 3.80%, and 3.61% compared to CACCD-GOADL, CytoBrain, CYENET, and SMOTETOMEK, respectively.</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

CERVIC-NET: Cervical Cancer Classification Using Dual-stream Capsule Pyramid Network with Multi-modality Images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月30日

查看全文
  • R. Sivaguru,
  • R. Thangarajan

摘要

Cervical cancer (CC) is the most prevalent cancer in women; it is a serious risk to the health of women. The colposcopy procedure is an essential part of CC prevention. However, existing methods often struggle with limited feature representation, poor multimodal fusion, and low accuracy in early-stage CC classification. To overcome these challenges, a novel CERVIC-NET is proposed for accurate classification and staging of CC using multi-modality images. The patient undergoes HPV DNA testing, where a negative result requires no further action, while a positive result leads to clinical screening for detailed cervical examination. The input both colposcopy and Pap smear images are pre-processed using Anisotropic Guided Filter (AGF) is utilized to enhance image clarity and suppress noise. The proposed Dual-Stream Capsule Pyramid Network (DSCP-NET) is employed to extract spatial and hierarchical features from both imaging modalities. The extracted features are fused to capture multimodal correlations and fed into Gated Multi-Layer Perceptron (GMLP) for accurately classify Negative for Squamous Cell Carcinoma (SCC), Low-Grade Squamous Intraepithelial Lesion (LSIL), High-Grade Squamous Intraepithelial Lesion (HSIL), and Intraepithelial Lesion or Malignancy (NILM). The effectiveness of the CERVIC-NET is evaluated utilizing recall, accuracy, precision, specificity, and F1-score. The overall accuracy and precision of the CERVIC-NET are 98.50% and 98.16%, respectively. The CERVIC-NET improves an overall accuracy by 9.62%, 4.86%, 3.80%, and 3.61% compared to CACCD-GOADL, CytoBrain, CYENET, and SMOTETOMEK, respectively.