A Study on Staging Cystic Echinococcosis Using Machine Learning Methods.

Abstract

Cystic echinococcosis (CE) is a chronic parasitic disease characterized by slow progression and non-specific clinical symptoms, often leading to delayed diagnosis and treatment. Early and precise diagnosis is crucial for effective treatment, particularly considering the five stages of CE as outlined by the World Health Organization (WHO). This study explores the development of an advanced system that leverages artificial intelligence (AI) and machine learning (ML) techniques to classify CE cysts into stages using various imaging modalities, including computed tomography (CT), ultrasound (US), and magnetic resonance imaging (MRI). A total of ten ML algorithms were evaluated across these datasets, using performance metrics such as accuracy, precision, recall (sensitivity), specificity, and F1 score. These metrics offer diverse criteria for assessing model performance. To address this, we propose a normalization and scoring technique that consolidates all metrics into a final score, allowing for the identification of the best model that meets the desired criteria for CE cyst classification. The experimental results demonstrate that hybrid models, such as CNN+ResNet and Inception+ResNet, consistently outperformed other models across all three datasets. Specifically, CNN+ResNet, selected as the best model, achieved 97.55% accuracy on CT images, 93.99% accuracy on US images, and 100% accuracy on MRI images. This research underscores the potential of hybrid and pre-trained models in advancing medical image classification, providing a promising approach to improving the differential diagnosis of CE disease.