Cervical Cancer Severity Characterization Using Machine Learning Techniques

Cervical cancer ranks as the fourth most prevalent cancer globally, particularly affecting developing countries. Early diagnosis can significantly enhance patient clinical management. However, there is a critical shortage of qualified cytotechnicians compared to the number of individuals requiring diagnosis. Computer aided diagnostic systems can substantially improve the accuracy, reliability, speed, and cost-effectiveness of diagnoses. Traditional machine learning diagnostic systems function similarly to cytopathologists, relying on handcrafted morphological features such as nucleus area to determine cell malignancy.

This study aims to develop machine learning algorithms for the automatic identification of cervical cancer cells based on the percentage of nucleus area affected and to quantify different cervical cancer stages. A comparison of various machine learning models such as support vector machine (SVM), random forest, and k-nearest neighbor is carried out to characterize cervical cancer cells.

The highest performance result of 91.80% is achieved using SVM classifier trained with shape features. This proposed machine learning framework assists radiologists in diagnosing the stage of cervical cancer.