Photo Credit: Marco Marca
A machine learning algorithm designed to predict MASH using clinical data and blood parameters led to earlier and more accurate diagnoses, which could potentially improve clinical outcomes and disease burden, according to a study published in Scientific Reports. Amir Reza Naderi Yaghouti and colleagues used predictive features from a dataset of 181 patients to train various machine learning classifiers, including support vector machine, random forest, AdaBoost, LightGBM, and XGBoost. The random forest model, combined with sequential forward selection and 10 features, demonstrated the best performance (accuracy=81.32%±6.43%; sensitivity=86.04%±6.21%; specificity=70.49%±8.12%; precision=81.59%±6.23%; and F1-score=83.75%±6.23%). The outcomes of the study underscore machine learning’s potential to augment and potentially replace invasive diagnostic procedures like liver biopsy, which is considered the gold standard for confirming MASH.
