Scholarlink

Purpose <p>The primary aim of this study is to develop a cutting-edge automated approach for detecting and classifying uterine myomas in medical images through deep learning. The research seeks to determine whether the U-Net + + architecture can accurately and efficiently identify and classify myomas, offering a dependable and advanced tool for clinical diagnostics.</p> Methods <p>The proposed method employs a U-Net + + architecture for myoma detection and classification. Preprocessing includes image normalization, resizing to 256 × 256 pixels, and contrast enhancement using a Gradient-Aware Contrast Limited Adaptive Histogram Equalization (GA-CLAHE) technique. The model is trained on a curated dataset, and its predictions are post-processed to crop and highlight the detected regions of interest accurately. The entire approach is validated against ground truth annotations to ensure reliable performance.</p> Results <p>The AG-CLAHE-UNET + + method achieved an accuracy of 0.99, with a sensitivity of 0.98 and a specificity of 0.98, demonstrating its effectiveness in detecting and classifying uterine myomas.</p> Conclusion <p>This study presents a comprehensive method for automated uterine myoma detection using U-Net++. The approach offers a robust and efficient tool for clinical use, with significant implications for improving diagnostic accuracy and efficiency.</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

Myoma detection from laparoscopy surgery videos using multilayer UNET + + model

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Lohith Reddy Kalluru,
  • Sohit Reddy Kalluru,
  • Srikanth Cherukuvada,
  • Naga Padmaja Jagini

摘要

Purpose

The primary aim of this study is to develop a cutting-edge automated approach for detecting and classifying uterine myomas in medical images through deep learning. The research seeks to determine whether the U-Net + + architecture can accurately and efficiently identify and classify myomas, offering a dependable and advanced tool for clinical diagnostics.

Methods

The proposed method employs a U-Net + + architecture for myoma detection and classification. Preprocessing includes image normalization, resizing to 256 × 256 pixels, and contrast enhancement using a Gradient-Aware Contrast Limited Adaptive Histogram Equalization (GA-CLAHE) technique. The model is trained on a curated dataset, and its predictions are post-processed to crop and highlight the detected regions of interest accurately. The entire approach is validated against ground truth annotations to ensure reliable performance.

Results

The AG-CLAHE-UNET + + method achieved an accuracy of 0.99, with a sensitivity of 0.98 and a specificity of 0.98, demonstrating its effectiveness in detecting and classifying uterine myomas.

Conclusion

This study presents a comprehensive method for automated uterine myoma detection using U-Net++. The approach offers a robust and efficient tool for clinical use, with significant implications for improving diagnostic accuracy and efficiency.