- Linus Mathias1, Giliki Jerison2 and Ishaku Kpapisom3
- DOI: 10.5281/zenodo.22091730
- SSR Journal of Artificial Intelligence (SSRJAI)
Nonalcoholic Fatty Liver Disease (NAFLD)
is one of the most common chronic liver diseases worldwide and is strongly
associated with obesity, type 2 diabetes mellitus, metabolic syndrome, and
other cardiovascular risk factors. If left undiagnosed, NAFLD may progress to
nonalcoholic steatohepatitis (NASH), liver fibrosis, cirrhosis, and
hepatocellular carcinoma, making early detection essential for effective
clinical management. Although magnetic resonance-based fat quantification is
regarded as the noninvasive reference standard for assessing hepatic steatosis,
its high cost, limited accessibility, and lengthy examination time restrict its
widespread use, particularly in low-resource healthcare settings. Ultrasound
imaging provides an affordable, safe, and widely available alternative;
however, its diagnostic accuracy is often affected by operator dependency and
subjective interpretation. This study investigates the feasibility of
integrating artificial intelligence (AI) with ultrasound imaging to improve the
accuracy and reliability of NAFLD detection. The proposed framework employs
advanced deep learning algorithms, including convolutional neural networks
(CNNs) and transfer learning, to automatically extract and classify
discriminative imaging features associated with hepatic fat accumulation. To
enhance model robustness and generalization, the training dataset was expanded
to include a larger and more diverse patient population. Furthermore,
multimodal data comprising ultrasound images, demographic information, clinical
history, and laboratory findings were integrated to improve diagnostic
performance and support more informed clinical decision-making. Model
performance was assessed using accuracy, precision, recall, F1-score, receiver
operating characteristic–area under the curve (ROC-AUC), k-fold
cross-validation, and external validation to ensure reliability and clinical
applicability. The findings indicate that AI-enhanced ultrasound can
significantly improve diagnostic consistency, reduce observer variability, and
provide a cost-effective, noninvasive solution for routine NAFLD screening and
early detection. The proposed approach demonstrates strong potential for
supporting clinical decision-making and improving patient outcomes, particularly
in resource-constrained healthcare environments where access to advanced
imaging modalities remains limited.
Keywords: Nonalcoholic
Fatty Liver Disease (NAFLD), Artificial Intelligence, Ultrasound Imaging, Deep
Learning, Convolutional Neural Networks, Transfer Learning, Multimodal Data,
Early Detection.
