Scholarlink

<p>Oral cancer remains a critical health issue because it is often identified only in its later stages, which severely affects survival outcomes. Conventional diagnostic techniques are not only invasive and resource-intensive but also rely heavily on manual judgment, which can cause delays and diagnostic inaccuracies. Hence, an efficient model named Goat Orangutan Optimization Algorithm based Convolutional Extreme Gradient Boosting (GOOA_ConvXGB) is designed for detecting oral cancer using multimodal data. The multimodal dataset includes histopathological images, sociodemographic data, and clinical data. Initially, the Mean-Shift Filter is used to denoise a histopathological image. Blood cell segmentation is performed by the Swin-Unet. Then, Haralick features and Pairwise Rotation Invariant Co-Occurrence Local Binary Pattern (PRICoLBP) are extracted. Thereafter, sociodemographic data and clinical data are fed for data normalization using Hyperbolic Tangent Function-Based Normalization. Finally, extracted features from histopathological images, along with normalized sociodemographic and clinical data, are aggregated to perform detection using Convolutional eXtreme Gradient Boosting (ConvXGB). Moreover, ConvXGB is tuned using GOOA, which is the combination of the Goat optimization Algorithm (GOA) and the Orangutan Optimization Algorithm (OOA). Furthermore, GOOA_ConvXGB has attained optimal results of 96.347% of accuracy, 96.811% of True Negative Rate (TNR), 95.861% of True Positive Rate (TPR), 95.656% of F1-score and 95.452% of 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

Goat Orangutan Optimization Algorithm-based Convolutional Extreme Gradient Boosting for Oral Cancer Detection using Multimodal Data

  • 期刊论文

  • 发表日期：

    2026年3月15日

查看全文
  • Pradeep Chauhan,
  • Raju Ranjan

摘要

Oral cancer remains a critical health issue because it is often identified only in its later stages, which severely affects survival outcomes. Conventional diagnostic techniques are not only invasive and resource-intensive but also rely heavily on manual judgment, which can cause delays and diagnostic inaccuracies. Hence, an efficient model named Goat Orangutan Optimization Algorithm based Convolutional Extreme Gradient Boosting (GOOA_ConvXGB) is designed for detecting oral cancer using multimodal data. The multimodal dataset includes histopathological images, sociodemographic data, and clinical data. Initially, the Mean-Shift Filter is used to denoise a histopathological image. Blood cell segmentation is performed by the Swin-Unet. Then, Haralick features and Pairwise Rotation Invariant Co-Occurrence Local Binary Pattern (PRICoLBP) are extracted. Thereafter, sociodemographic data and clinical data are fed for data normalization using Hyperbolic Tangent Function-Based Normalization. Finally, extracted features from histopathological images, along with normalized sociodemographic and clinical data, are aggregated to perform detection using Convolutional eXtreme Gradient Boosting (ConvXGB). Moreover, ConvXGB is tuned using GOOA, which is the combination of the Goat optimization Algorithm (GOA) and the Orangutan Optimization Algorithm (OOA). Furthermore, GOOA_ConvXGB has attained optimal results of 96.347% of accuracy, 96.811% of True Negative Rate (TNR), 95.861% of True Positive Rate (TPR), 95.656% of F1-score and 95.452% of precision.