Texture Features-Based Machine Learning for Classification of Cervical Cell Microscopic Images

Machine learning has been developed in biomedical science as a clinical decision-support technique. It can automatically recognize patterns in a given dataset to perform predictions and data classification. It has been widely applied to detect abnormalities in various diseases by leveraging texture-based features. This study utilized a Histogram of Oriented Gradients (HOG) as a feature extraction method and a Support Vector Machine (SVM) for classification in detecting the level of cancer abnormalities in cervical cells. The dataset of cervical cell microscopic images consisted of 972 images for training and 108 for testing. Among the models used, the Medium Gaussian SVM achieved the best performance, with a training accuracy of 73.3% and the fastest training time of 4.123 seconds. Whereas, the best performance matrix of testing results achieved 81.10% of accuracy. These results highlight the potential of the proposed method in improving cervical cancer detection through automated and efficient classification techniques.