Scholarlink

<p>This research pioneers breast cancer detection with the “DeepRNN-CNN Ensemble,” an advanced methodology integrating sophisticated techniques at multiple stages for precision. The proposed approach commences with comprehensive preprocessing steps, including color normalization through intensity normalization, stain normalization employing color deconvolution, Gaussian blurring for noise reduction, and a novel adaptive histogram equalization method for image enhancement. The segmentation phase utilizes the optimized Deep Lab technique, tailored for superior segmentation accuracy. Feature extraction encompasses a multi-faceted approach, extracting traditional handcrafted features such as Local Binary Patterns, Haralick features, and Zernike moments. Additionally, spatial-temporal features are derived based on the Discrete Space-Time Pyramid. Deep features are harnessed using the Visual Geometry Group (VGG), allowing for the extraction of intricate patterns from histopathology images through fine-tuning or feature extraction. The innovation extends to feature selection, employing a Hybrid Optimization Algorithm that combines the Walrus. Optimization Algorithm and the red-tailed hawk algorithm to improve the discriminative power of selected features. The heart of the suggested methodology lies in the DeepRNN-CNN Ensemble-based detection architecture, integrating Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Bidirectional Long Short-Term Memory (Bi-LSTM) components. The cascading layers, from global average pooling to dense layers and Bi-LSTM-RNN, utilize a combination of recurrent and convolutional cells for optimal information processing. The integration of RNN instead of GRU, attention mechanisms, and the outcome prediction showcase the adaptability and efficiency of the ensemble. This cutting-edge methodology not only pushes the boundaries of breast cancer detection accuracy but also provides a versatile framework for advancing medical image analysis for enhanced diagnostic precision and treatment. This method excels in breast cancer detection, combining refined segmentation, hybrid feature selection, and a sophisticated ensemble architecture, marking a state-of-the-art advancement. The proposed method outperformed benchmark models in experimental validation, achieving 98.75% accuracy, high sensitivity, specificity, precision, and an excellent Matthews Correlation Coefficient (0.9876) while maintaining a low rate of false positives and false negatives.</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

DeepRNN-CNN ensemble with optimized segmentation and hybrid feature selection for improved breast cancer detection

  • 期刊论文

  • 发表日期：

    2026年2月17日

查看全文
  • Priyanka Dashrathsinh Puvar,
  • Bhagirath Parshuram Prajapati

摘要

This research pioneers breast cancer detection with the “DeepRNN-CNN Ensemble,” an advanced methodology integrating sophisticated techniques at multiple stages for precision. The proposed approach commences with comprehensive preprocessing steps, including color normalization through intensity normalization, stain normalization employing color deconvolution, Gaussian blurring for noise reduction, and a novel adaptive histogram equalization method for image enhancement. The segmentation phase utilizes the optimized Deep Lab technique, tailored for superior segmentation accuracy. Feature extraction encompasses a multi-faceted approach, extracting traditional handcrafted features such as Local Binary Patterns, Haralick features, and Zernike moments. Additionally, spatial-temporal features are derived based on the Discrete Space-Time Pyramid. Deep features are harnessed using the Visual Geometry Group (VGG), allowing for the extraction of intricate patterns from histopathology images through fine-tuning or feature extraction. The innovation extends to feature selection, employing a Hybrid Optimization Algorithm that combines the Walrus. Optimization Algorithm and the red-tailed hawk algorithm to improve the discriminative power of selected features. The heart of the suggested methodology lies in the DeepRNN-CNN Ensemble-based detection architecture, integrating Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Bidirectional Long Short-Term Memory (Bi-LSTM) components. The cascading layers, from global average pooling to dense layers and Bi-LSTM-RNN, utilize a combination of recurrent and convolutional cells for optimal information processing. The integration of RNN instead of GRU, attention mechanisms, and the outcome prediction showcase the adaptability and efficiency of the ensemble. This cutting-edge methodology not only pushes the boundaries of breast cancer detection accuracy but also provides a versatile framework for advancing medical image analysis for enhanced diagnostic precision and treatment. This method excels in breast cancer detection, combining refined segmentation, hybrid feature selection, and a sophisticated ensemble architecture, marking a state-of-the-art advancement. The proposed method outperformed benchmark models in experimental validation, achieving 98.75% accuracy, high sensitivity, specificity, precision, and an excellent Matthews Correlation Coefficient (0.9876) while maintaining a low rate of false positives and false negatives.