Scholarlink

<p>One of the most prevalent and severe tumors in women, breast cancer, remains a major global health issue despite a notable increase in incidence over the last ten years. It is the second leading cause of cancer-related death among women. Identifying breast cancer in its early stages has the potential to save lives; however, current screening techniques for the illness require several laboratory procedures involving medical experts. Automated solutions with rapid and reliable diagnostic capabilities are needed to minimize human error and expedite breast cancer diagnosis. The projected accuracy of cancer diagnosis remains far from matching the precision offered by existing approaches, even with the research on automated systems for the disease being studied. This work suggests a long short-term memory-based modified Gorilla troop optimization (LSTM-MGTO) method for breast cancer classification in order to address these issues. The Mastectomy Koibra Dataset (BCCD) and Wisconsin Diagnostic Mastectomy (WDBC) datasets were used to test the suggested methods. First, the proposed system employs contrast-limited adaptive histogram equalization (CLAHE) to enhance the quality of digital mammograms.. Furthermore, employ a semantic deep learning (SDL) model to extract features. After the feature selection process, a recursive feature elimination technique was implemented to determine the crucial WDBC and BCCD characteristics that are relevant to breast cancer detection. Moreover, recommend a modified U-Net architecture for partitioning in both unmapped and guided contexts. The experimental findings indicate that the newly developed partitioning model surpasses existing advanced techniques, yielding superior results in both Dice and IoU score evaluations. On the WDBC and BCCD datasets, the suggested U-Net segmentation produces maximum Dice scores of 97.65% and 96.24%, respectively. Additionally, the model obtained the greatest IoU scores of 95.43% and 90.65% on the WDBC and BCCD datasets, respectively, according to the experimental findings.</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

LSTM-MGTO: a novel early breast cancer detection using long short term memory based modified gorilla troops optimization algorithm

  • 期刊论文

  • 发表日期：

    2025年3月17日

查看全文
  • D. Umamaheswari,
  • M. Kannan,
  • I. Priya Stella Mary,
  • D. Juliet Rozario,
  • P. Margaret Savitha,
  • B. Manimekala

摘要

One of the most prevalent and severe tumors in women, breast cancer, remains a major global health issue despite a notable increase in incidence over the last ten years. It is the second leading cause of cancer-related death among women. Identifying breast cancer in its early stages has the potential to save lives; however, current screening techniques for the illness require several laboratory procedures involving medical experts. Automated solutions with rapid and reliable diagnostic capabilities are needed to minimize human error and expedite breast cancer diagnosis. The projected accuracy of cancer diagnosis remains far from matching the precision offered by existing approaches, even with the research on automated systems for the disease being studied. This work suggests a long short-term memory-based modified Gorilla troop optimization (LSTM-MGTO) method for breast cancer classification in order to address these issues. The Mastectomy Koibra Dataset (BCCD) and Wisconsin Diagnostic Mastectomy (WDBC) datasets were used to test the suggested methods. First, the proposed system employs contrast-limited adaptive histogram equalization (CLAHE) to enhance the quality of digital mammograms.. Furthermore, employ a semantic deep learning (SDL) model to extract features. After the feature selection process, a recursive feature elimination technique was implemented to determine the crucial WDBC and BCCD characteristics that are relevant to breast cancer detection. Moreover, recommend a modified U-Net architecture for partitioning in both unmapped and guided contexts. The experimental findings indicate that the newly developed partitioning model surpasses existing advanced techniques, yielding superior results in both Dice and IoU score evaluations. On the WDBC and BCCD datasets, the suggested U-Net segmentation produces maximum Dice scores of 97.65% and 96.24%, respectively. Additionally, the model obtained the greatest IoU scores of 95.43% and 90.65% on the WDBC and BCCD datasets, respectively, according to the experimental findings.