Accurate Classification of Ovarian Cyst Using Topic Modeling

Ovaries are an essential female organ in the reproductive and endocrine system, responsible for healthy development and foster fertility. Occasionally, they are affected by vesicles that are filled with fluid and are referred to as cyst. Most of the ovarian cysts are benign in nature, but in few cases, they can be harmful for the patient and need to be treated at an early phase. Globally, ovarian cyst is a prevalent problem among females. The identification and classification of ovarian cyst is a major task in medical data analysis for healthcare providers. Manual classification of ovarian cyst may lead to delay in the diagnosis of the patient. In this paper, three Topic Modeling algorithms are utilized to classify ultrasound textual reports into cystic and normal ovary reports. The results of the algorithms are analyzed on the basis of six evaluation metrics like, accuracy, precision, recall, etc. Latent Dirichlet Allocation with accuracy of 90.71% is proven to be a better algorithm for ovarian cyst classification compared to Non-negative Matrix Factorization and Latent Semantic Analysis with accuracies of 83.79% and 90.34% respectively.