Alzakari Sarah A, Alhussan Amel Ali, Towfek S K, Metwally Marwa, Salem Dina Ahmed
Department of Computer Sciences, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, P.O. Box 84428, Riyadh 11671, Saudi Arabia.
Computer Science and Intelligent Systems Research Center, Blacksburg, VA 24060, USA.
Bioengineering (Basel). 2025 Apr 30;12(5):478. doi: 10.3390/bioengineering12050478.
More than 85% of the world's cervical cancer fatalities occur in less-developed nations, causing early mortality among women. In this paper, we propose a novel approach for the early classification of cervical cancer based on a new feature selection algorithm and classification method. The new feature selection algorithm is based on a hybrid of the Waterwheel Plant Algorithm and Particle Swarm Optimization algorithms, and bWWPAPSO denotes it. Meanwhile, the new classification method is based on optimizing the parameters of a multilayer perceptron neural network (WWPAPSO+MLP). A publicly available dataset is employed to verify the effectiveness of the proposed approach. Due to this dataset's imbalance and missing values, it is preprocessed and balanced using SMOTETomek, where undersampling and oversampling were utilized. The usefulness of class imbalance and feature selection based on the classifier's specificity, sensitivity, and accuracy has been demonstrated by way of a comparative study of the proposed methodology that has been carried out. WWPAPSO+MLP achieves superior performance, with an accuracy of 97.3% and a sensitivity of 98.8%. On the other hand, several statistical tests were conducted, including the Wilcoxon signed rank test and analysis of variance (ANOVA) to confirm the effectiveness and superiority of the proposed approach.
全球超过85%的宫颈癌死亡病例发生在欠发达国家，导致女性过早死亡。在本文中，我们提出了一种基于新的特征选择算法和分类方法的宫颈癌早期分类新方法。新的特征选择算法基于水车植物算法和粒子群优化算法的混合，并用bWWPAPSO表示。同时，新的分类方法基于对多层感知器神经网络参数的优化（WWPAPSO+MLP）。使用一个公开可用的数据集来验证所提方法的有效性。由于该数据集存在不平衡和缺失值的问题，使用SMOTETomek对其进行预处理和平衡，其中采用了欠采样和过采样。通过对所提方法进行比较研究，证明了基于分类器特异性、敏感性和准确性的类别不平衡和特征选择的有用性。WWPAPSO+MLP表现出卓越的性能，准确率为97.3%，敏感性为98.8%。另一方面，进行了多项统计检验，包括威尔科克森符号秩检验和方差分析（ANOVA），以确认所提方法的有效性和优越性。
