Doctors often struggle to see early signs of fatty liver disease before it becomes serious. A new look at computer tools shows they might help solve this problem. Researchers combined results from 106 different studies to test how well these programs work.
The analysis focused on two types of smart software. Machine learning models found a score of 0.833 for spotting fatty liver disease. Deep learning models did even better with a score of 0.841. These numbers measure how often the computer gets the diagnosis right compared to a human expert.
The tools also worked well for finding liver scarring. One specific machine learning program called CatBoost reached a perfect score of 0.960. Another deep learning program called ResNet50 reached 0.960 as well. These results suggest the technology is ready for real-world use.
This review did not report any safety issues because these are computer programs, not drugs. The findings come from a large group of studies, which makes the results more reliable. While more research is always good, these tools offer a clear path forward for better liver care.