Mekala Sandhya, S Phani Kumar
GITAM School of Technology, GITAM Deemed To Be University, Rudraram, Patancheru Mandal, Hyderabad, Telangana, 502329, India.
Sci Rep. 2025 Jun 5;15(1):19787. doi: 10.1038/s41598-025-00512-6.
The pancreas is a gland in the abdomen that helps to produce hormones and digest food. The irregular development of tissues in the pancreas is termed as pancreatic cancer. Identification of pancreatic tumors early is significant for enhancing survival rate and providing appropriate treatment. Thus, an efficient Secretary Wolf Bird Optimization (SeWBO)_Efficient DenseNet is presented for pancreatic tumor detection using Computed Tomography (CT) scans. Firstly, the input pancreatic CT image is accumulated from a database and subjected to image preprocessing using a bilateral filter. After this, lesion is segmented by utilizing Parallel Reverse Attention Network (PraNet), and hyperparameters of PraNet are enhanced by using the proposed SeWBO. The SeWBO is designed by incorporating Wolf Bird Optimization (WBO) and the Secretary Bird Optimization Algorithm (SBOA). Then, features like Complete Local Binary Pattern (CLBP) with Discrete Wavelet Transformation (DWT), statistical features, and Shape Local Binary Texture (SLBT) are extracted. Finally, pancreatic tumor detection is performed by SeWBO_Efficient DenseNet. Here, Efficient DenseNet is developed by combining EfficientNet and DenseNet. Moreover, the proposed SeWBO_Efficient DenseNet achieves better True Negative Rate (TNR), accuracy, and True Positive Rate (TPR), of 93.596%, 94.635%, and 92.579%.
胰腺是腹部的一个腺体，有助于产生激素和消化食物。胰腺组织的异常发育被称为胰腺癌。早期识别胰腺肿瘤对于提高生存率和提供适当治疗具有重要意义。因此，提出了一种高效的秘书狼鸟优化（SeWBO）_高效密集连接网络（Efficient DenseNet），用于通过计算机断层扫描（CT）扫描检测胰腺肿瘤。首先，从数据库中积累输入的胰腺CT图像，并使用双边滤波器对其进行图像预处理。在此之后，利用并行反向注意力网络（PraNet）对病变进行分割，并使用所提出的SeWBO增强PraNet的超参数。SeWBO是通过结合狼鸟优化（WBO）和秘书鸟优化算法（SBOA）设计的。然后，提取诸如带有离散小波变换（DWT）的完整局部二值模式（CLBP）、统计特征和形状局部二值纹理（SLBT）等特征。最后，由SeWBO_高效密集连接网络进行胰腺肿瘤检测。在这里，高效密集连接网络是通过结合高效网络（EfficientNet）和密集连接网络（DenseNet）开发的。此外，所提出的SeWBO_高效密集连接网络实现了更好的真阴性率（TNR）、准确率和真阳性率（TPR），分别为93.596%、94.635%和92.579%。
