论文标题
对PAP涂片图像分类的深度学习和传统机器学习技术的比较
Comparison of Deep Learning and Traditional Machine Learning Techniques for Classification of Pap Smear Images
论文作者
论文摘要
通过使用HERLEV数据集结果的PAP涂片图像，对机器和深度学习技术进行了一项针对正常和异常宫颈细胞分类的研究。该数据集包括917张图像和7个不同类别。这项研究中使用的所有技术都是通过在Tensorflow内与Scikit-Learn和Keras库的Google Colab平台建模的。在第一项研究中，使用了传统的机器学习方法，例如逻辑回归，K-Nearest邻居（KNN），支持向量机（SVM），决策树，随机森林和极端梯度增强（XGBoost）（XGBoost），并相互比较以发现正常和异常宫颈细胞。 XGBOOST和KNN分类器的精度为85％，可以观察到更好的结果。在第二项研究中，使用基于卷积神经网络（CNN）的深度学习模型用于同一数据集。因此，培训和测试数据集获得了99％和93％的精度。在此模型中，在计算时间20分钟内具有这些精度需要50个时期。
A comprehensive study on machine and deep learning techniques for classification of normal and abnormal cervical cells by using pap smear images from Herlev dataset results are presented. This dataset includes 917 images and 7 different classes. All techniques used in this study are modeled by using Google Colab platform with scikit-learn and Keras library inside TensorFlow. In the first study, traditional machine learning methods such as logistic regression, k-Nearest Neighbors (kNN), Support Vector Machine (SVM), Decision Tree, Random Forest and eXtreme Gradient Boosting (XGBoost) are used and compared with each other to find binary classification as normal and abnormal cervical cells. Better results are observed by XGBoost and kNN classifiers among the others with an accuracy of 85%. In the second study, a deep learning model based on Convolutional Neural Network(CNN) is used for the same dataset. Accordingly, accuracies of 99% and 93% are obtained for the training and the test dataset, respectively. In this model, it takes 50 epochs to have these accuracies within 20 minutes of computational time.
