Scholarlink

<p>Cancer is possibly the foremost disease which poses a greater risk for people in low and middle-income populations among the countries. Treatments are costly, and they corroborate the need for effective screening and more risk factors that have been identified as smoking, above 45&#xa0;years of age, economic status, and environmental exposures. As per the objective of improving the early detection of lung cancer, this study evaluates deep learning models that include a combination of convolutional neural networks (CNNs) like DenseNet201, ResNet152V2, and MobileNetV2, with bidirectional long short-term memory (Bi-LSTM) layers, performed with Adam, SGD, and RMSprop optimizers. When Bi-LSTM was absent, the standard CNNs performed adequately well, with SGD being competitive. The additional use of Bi-LSTM raised performance levels, where ResNet152V2-Bi-LSTM excelled above all possible architectural setups, particularly regarding SGD optimization. AUC-ROC analysis confirmed its ability to capture discriminative features; further metrics of an accuracy-recall-precision-MCC-F1-score established it as having a performance edge over all other models. The findings bring home the superior accuracy, recall, and precision that ResNet152V2-Bi-LSTM offers-spell critical importance when it comes to lung cancer detection, and reminds one to keep in mind the ends of model and optimizer choices regarding medical imaging. Built and optimized using SGD, it will offer promising leaps in fast disease diagnosis and valuable forms of support in making clinical decisions and patient care. Such studies show transformational roles for customized deep learning approaches in improving diagnostic and clinical performance.</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

Enhanced Lung Cancer Detection Using Deep Learning for Chest Radiograph Classification: A Comparison of CNNs with RNN Components

  • 期刊论文

  • 发表日期：

    2025年5月23日

查看全文
  • Y. Arockia Raj,
  • Pamarthi Sunitha,
  • M. Anandan,
  • Yousef Farhaoui

摘要

Cancer is possibly the foremost disease which poses a greater risk for people in low and middle-income populations among the countries. Treatments are costly, and they corroborate the need for effective screening and more risk factors that have been identified as smoking, above 45 years of age, economic status, and environmental exposures. As per the objective of improving the early detection of lung cancer, this study evaluates deep learning models that include a combination of convolutional neural networks (CNNs) like DenseNet201, ResNet152V2, and MobileNetV2, with bidirectional long short-term memory (Bi-LSTM) layers, performed with Adam, SGD, and RMSprop optimizers. When Bi-LSTM was absent, the standard CNNs performed adequately well, with SGD being competitive. The additional use of Bi-LSTM raised performance levels, where ResNet152V2-Bi-LSTM excelled above all possible architectural setups, particularly regarding SGD optimization. AUC-ROC analysis confirmed its ability to capture discriminative features; further metrics of an accuracy-recall-precision-MCC-F1-score established it as having a performance edge over all other models. The findings bring home the superior accuracy, recall, and precision that ResNet152V2-Bi-LSTM offers-spell critical importance when it comes to lung cancer detection, and reminds one to keep in mind the ends of model and optimizer choices regarding medical imaging. Built and optimized using SGD, it will offer promising leaps in fast disease diagnosis and valuable forms of support in making clinical decisions and patient care. Such studies show transformational roles for customized deep learning approaches in improving diagnostic and clinical performance.