Scholarlink

Despite advances in various screened and diagnostic techniques, breast cancer diagnosis and prognosis in the general population, and especially in women with dense breast tissue, still consists of a considerable challenge. Such restriction may lead to extended shifts in the diagnosis and in some cases avoidable actions, which signifies the importance of more efficient mechanisms. This research addresses these gaps by using synthetic data generated with Generative Adversarial Networks (GANs) to enhance breast cancer classification, addressing the problem of class imbalance—a frequent drawback of medical datasets. Synthetic data generated from GANs adds realistic samples into the training set of the model, thereby strengthening its ability to differentiate between malignant and benign cases more accurately. Using a convolutional neural network (CNN) for feature extraction and expanding the training dataset by using GANs, the model was able to achieve excellent scores in validation accuracy of 99.17%, precision of 98.84%, recall of 100%, MCC of 0.9801, and ROC AUC of 0.9998. The combination of such accuracy with inherent features of explainability allow for the oversight by clinicians and espousal of clinical decisions consequently minimizing false negatives and unnecessary biopsies. Hence, the system improves not only the patient outcomes but also reduces the load on the health-care systems.

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

Synergizing Generative Adversarial Network-Driven Synthetic Data Pipelines with Deep Neural Networks for Enhanced Breast Cancer Diagnosis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Garima Shukla,
  • Vanshaj Awasthi,
  • Sakshi Nipane,
  • Dolly Sharma,
  • Sofia Singh

摘要

Despite advances in various screened and diagnostic techniques, breast cancer diagnosis and prognosis in the general population, and especially in women with dense breast tissue, still consists of a considerable challenge. Such restriction may lead to extended shifts in the diagnosis and in some cases avoidable actions, which signifies the importance of more efficient mechanisms. This research addresses these gaps by using synthetic data generated with Generative Adversarial Networks (GANs) to enhance breast cancer classification, addressing the problem of class imbalance—a frequent drawback of medical datasets. Synthetic data generated from GANs adds realistic samples into the training set of the model, thereby strengthening its ability to differentiate between malignant and benign cases more accurately. Using a convolutional neural network (CNN) for feature extraction and expanding the training dataset by using GANs, the model was able to achieve excellent scores in validation accuracy of 99.17%, precision of 98.84%, recall of 100%, MCC of 0.9801, and ROC AUC of 0.9998. The combination of such accuracy with inherent features of explainability allow for the oversight by clinicians and espousal of clinical decisions consequently minimizing false negatives and unnecessary biopsies. Hence, the system improves not only the patient outcomes but also reduces the load on the health-care systems.