Automatic Classification of HeLa Cell Images Using Machine Learning Techniques

Abstract

This thesis investigates the automatic classification of HeLa cell images using machine learning techniques, with a focus on comparing Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs). A CNN model is developed to learn features directly from microscopy images, while an SVM classifier is applied to the extracted features to evaluate a hybrid approach. The dataset consists of labeled HeLa cell images with variations in morphology and imaging conditions, making classification a challenging task. Experimental results show that while the CNN achieves moderate accuracy, the hybrid CNN–SVM approach significantly improves performance. The findings indicate that CNNs are effective feature extractors, but alternative classifiers such as SVMs can better utilize these features for classification. Overall, the study demonstrates that combining deep learning and traditional machine learning methods can enhance performance in medical image classification tasks with limited data.