High Accuracy Ultrasound Image-Based Deep Learning for Ovarian Cyst Classification

Abstract

Ovarian cysts present a significant health risk for women of all ages throughout the world. The identification and classification of ovarian cysts is one of the complicated medical tasks. Correct and precise identification of ovarian cysts can help patients for their speedy recovery and healthy well-being. Recently, Deep Learning models have become valuable tools for classification in medical imaging. This paper presents an improved Densely connected Network (DenseNet-169) to classify ovarian cysts with high efficacy. The effectiveness of the proposed model is assessed by evaluation indicators, such as accuracy, precision, f1-score and recall. DenseNet-169 has proved to be a better Deep Learning model for accurate and reliable categorization of ovarian cysts using ultrasound images with 100% as recall, 99.78% as accuracy, 99.58% as precision and 99.79% as F1-score. This paper marks a significant advancement in distinguishing cystic ovaries from normal ones and thereby, contributing to improved diagnostic capabilities in ovarian pathology.

Keywords: Deep learning, DenseNet-169, Ovarian cyst classification, Ultrasound images

View PDF