Enhancement of liver cirrhosis detection from ultrasonic images through clustering-based features and classifiers

Liver cirrhosis is ranked among the major chronic illnesses and is often present in millions of individuals around the world. It is caused by progressively developing fibrosis of the liver resulting in compromised function and ultimately, illness. Early detection and accurate diagnosis are arguably, key to effective treatment, avoiding the potential progression of disease and recovery of the patient. Conventional diagnostic approaches, including biochemical assessments and imaging practices, are accepted and known methods of diagnosis. Recent technological advances in machine learning and artificial intelligence have provided potential and probable alternatives to improving diagnostic care. This paper investigates and analyzes ultrasound images for liver cirrhosis in The Cancer Imaging Archive (TCIA) database and applies two feature extraction techniques, Fuzzy C-Means (FCM) and Possibilistic Fuzzy C-Means (PFCM). Eight models are used for classification, those are Gaussian Mixture Model (GMM), Particle Swarm Optimization algorithm (PSO), Cuckoo Search algorithm (CS), Artificial Algae Optimisation (AAO), Naïve Bayes Classifier (NBC) and a hybrid classification, PSO GMM, Cuckoo Search GMM (CS-GMM) AAO-GMM. The results of this study illustrate the good classification performance achieved with the FCM features using the bio-inspired AAO classifier (91.67%), and the highest classification performance achieved using the PFCM features plus the hybrid AAO-GMM classifier (99.03%). These findings exhibit the efficacy of hybrid machine learning algorithms in advancing liver cirrhosis prediction.