A NEW sequential diagnostic algorithm combining the Fibrosis-4 Index (FIB-4) and an ultrasound-based deep learning (DL) model has demonstrated improved accuracy and referral efficiency for identifying advanced liver fibrosis, a study reveals.
The research, conducted across three healthcare facilities between January 2014 and September 2022, aimed to address limitations in current non-invasive screening methods for patients with chronic liver disease. Using retrospective data, researchers developed a DL network, FIB-Net, to predict liver stiffness based on ultrasound images.
FIB-Net was designed to identify patients with a shear-wave elastography (SWE) value of 8.7 kPa or higher, indicative of advanced fibrosis. Researchers then tested two algorithms: a two-step approach (Two-step#1) combining FIB-4 and FIB-Net, and a three-step approach (Three-step#1), which added SWE measurements to the sequence. Both were assessed against established guidelines and standalone tests.
The study analysed 5,894 patients across training, validation, and test datasets. Results showed the Two-step#1 algorithm significantly outperformed FIB-4 alone, achieving higher specificity (79% vs 57%) and positive predictive value (PPV, 44% vs 32%), while reducing unnecessary referrals by 42%.
Similarly, the Three-step#1 algorithm excelled compared with the European Association for the Study of the Liver (EASL) guidelines, showing superior specificity (94% vs 88%) and PPV (73% vs 64%) and reducing unnecessary referrals by 35%.
Lead researchers highlighted the importance of such advancements in liver disease diagnostics, noting the inefficiencies of single-test approaches. “Combining FIB-4 with a DL model optimises both diagnostic accuracy and patient management, paving the way for better outcomes in advanced liver fibrosis screening.”
The findings underscore the potential for integrating artificial intelligence into clinical workflows, offering a more robust and streamlined approach to tackling liver fibrosis. This method could alleviate strain on healthcare systems by prioritising high-risk patients for further testing or treatment.
The study marks a significant step forward in the use of AI-enhanced tools for non-invasive disease management, with implications for improving care pathways for liver disease patients worldwide.
Reference
Chen L-D et al. US-based sequential algorithm integrating an ai model for advanced liver fibrosis screening. Radiology. 2024;311(1):e231461.
