Scholarlink

<p>Breast cancer remains a common and deadly disease that requires early and precise detection, especially in people with Phosphatase and Tensin Homolog (PTEN) Hamartoma Tumor Syndrome (PHTS). This research introduces a novel advanced hybrid diagnostic framework, Bayesian Asymmetric Quantized Neural Network optimized with the Quokka Swarm Optimization Algorithm (BAQNN-QSOA), to enhance the detection of BC in DDSM and DCE-MRI datasets. The proposed approach incorporates an Iterative Unsupervised Deep Bilateral Texture Filtering (IUDBTF) technique for noise suppression, followed by enhanced lesion segmentation through an Improved UNet architecture and Maximum-Entropy Regularized Decision Transformer (MERDT). For feature extraction, Multi-Discrete Wavelet Transform (MDWT) is employed to capture spectral-spatial characteristics. Lastly, the BAQNN model is optimized for the best classification accuracy using the Quokka Swarm Optimization Algorithm. With exceptional performance metrics, including Recall (0.986), Matthews Correlation Coefficient (MCC) (0.961), AUC (0.986), Specificity (0.986), F1-Score (0.986), Index of Jaccard (0.966), Cohen’s Kappa (0.959), Precision (0.986), Critical Success Index (0.965), and Accuracy (0.986) in terms of error rates and computing time, among other aspects, the proposed approach performs better than the most advanced models. The Friedman and Mann–Whitney U-tests were used to confirm the model’s robustness and dependability as well as the statistical significance of the findings. Compared to baseline methods including UNet + ResNet101, DeepLabV3 + VGG16, and Bi-GRU + EfficientNet, the BAQNN-QSOA framework consistently achieved higher classification accuracy and efficiency across both datasets. The study’s findings demonstrate that the proposed BAQNN-QSOA approach is a very reliable and promising method for diagnosing breast cancer in PHTS patients early on, providing better clinical decision support, decreased misdiagnosis, and increased precision.</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

Bayesian Asymmetric Quantized Neural Networks for MRI/Mammography-Based Breast Cancer Identification in PTEN Hamartoma Syndrome

  • 期刊论文

  • 发表日期：

    2025年8月4日

查看全文
  • A. Harshavardhan,
  • Afaque Alam,
  • R. Senthil Kumar,
  • Balasubbareddy Mallala

摘要

Breast cancer remains a common and deadly disease that requires early and precise detection, especially in people with Phosphatase and Tensin Homolog (PTEN) Hamartoma Tumor Syndrome (PHTS). This research introduces a novel advanced hybrid diagnostic framework, Bayesian Asymmetric Quantized Neural Network optimized with the Quokka Swarm Optimization Algorithm (BAQNN-QSOA), to enhance the detection of BC in DDSM and DCE-MRI datasets. The proposed approach incorporates an Iterative Unsupervised Deep Bilateral Texture Filtering (IUDBTF) technique for noise suppression, followed by enhanced lesion segmentation through an Improved UNet architecture and Maximum-Entropy Regularized Decision Transformer (MERDT). For feature extraction, Multi-Discrete Wavelet Transform (MDWT) is employed to capture spectral-spatial characteristics. Lastly, the BAQNN model is optimized for the best classification accuracy using the Quokka Swarm Optimization Algorithm. With exceptional performance metrics, including Recall (0.986), Matthews Correlation Coefficient (MCC) (0.961), AUC (0.986), Specificity (0.986), F1-Score (0.986), Index of Jaccard (0.966), Cohen’s Kappa (0.959), Precision (0.986), Critical Success Index (0.965), and Accuracy (0.986) in terms of error rates and computing time, among other aspects, the proposed approach performs better than the most advanced models. The Friedman and Mann–Whitney U-tests were used to confirm the model’s robustness and dependability as well as the statistical significance of the findings. Compared to baseline methods including UNet + ResNet101, DeepLabV3 + VGG16, and Bi-GRU + EfficientNet, the BAQNN-QSOA framework consistently achieved higher classification accuracy and efficiency across both datasets. The study’s findings demonstrate that the proposed BAQNN-QSOA approach is a very reliable and promising method for diagnosing breast cancer in PHTS patients early on, providing better clinical decision support, decreased misdiagnosis, and increased precision.