Machine learning to predict progression of non‐alcoholic fatty liver to non‐alcoholic steatohepatitis or fibrosis

Abstract

Background: Non-alcoholic fatty liver (NAFL) can progress to the severe subtype non-alcoholic steatohepatitis (NASH) and/or fibrosis, which are associated with increased morbidity, mortality, and healthcare costs. Current machine learning studies detect NASH; however, this study is unique in predicting the progression of NAFL patients to NASH or fibrosis. Aim: To utilize clinical information from NAFL-diagnosed patients to predict the likelihood of progression to NASH or fibrosis. Methods: Data were collected from electronic health records of patients receiving a first-time NAFL diagnosis. A gradient boosted machine learning algorithm (XGBoost) as well as logistic regression (LR) and multi-layer perceptron (MLP) models were developed. A five-fold cross-validation grid search was utilized for hyperparameter optimization of variables, including maximum tree depth, learning rate, and number of estimators. Predictions of patients likely to progress to NASH or fibrosis within 4 years of initial NAFL diagnosis were made using demographic features, vital signs, and laboratory measurements. Results: The XGBoost algorithm achieved area under the receiver operating characteristic (AUROC) values of 0.79 for prediction of progression to NASH and 0.87 for fibrosis on both hold-out and external validation test sets. The XGBoost algorithm outperformed the LR and MLP models for both NASH and fibrosis prediction on all metrics. Conclusion: It is possible to accurately identify newly diagnosed NAFL patients at high risk of progression to NASH or fibrosis. Early identification of these patients may allow for increased clinical monitoring, more aggressive preventative measures to slow the progression of NAFL and fibrosis, and efficient clinical trial enrollment.

An explainable machine learning model for prediction of high-risk nonalcoholic steatohepatitis

“…Although prior applications of traditional machine learning to detect MASLD, MASH and advance fibrosis have produced promising results, we showed that our explainable XGBoost MASLD model outperformed previous traditional ML models in detecting high-risk MASLD. Compared to Wu et al 19 , Docherty et al 4 , and Ghandian et al 5 , who reported sensitivities up to 0.82 and specificities up to 0.79, our model achieved comparable sensitivity (mean [95%], 0.71 [0.59–0.83]), but higher specificity (0.97 [0.96–0.98]), accuracy (0.95 [0.94–0.97]), and AUROC (0.95 [0.91–0.97]). Additionally, we report a good PPV of 0.59 [0.47–0.70] and excellent NPV of 0.98 [0.97–0.99].…”

“…Machine learning (ML) methods have shown promise in non-invasive diagnosis of MASH and advanced liver fibrosis using routine clinical data. A comparative study of various ML methods found that gradient boosting achieved the best area under the curve (AUC) scores for MASH and advanced fibrosis 4 , 5 . These ML models have the potential to improve MASH detection and diagnosis, ultimately leading to better management and outcomes for patients.…”

Prognostic Impact of Metabolic Syndrome and Steatotic Liver Disease in Hepatocellular Carcinoma Using Machine Learning Techniques

“…With disease progression, there is a parallel increase in the production of proinflammatory cytokines, such as tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), interleukin-32 (IL-32), or transforming growth factor-beta (TGF-β), which exceed the immune system’s control capacity through regulatory T cells [ 54 , 55 ]. This favors the onset of steatohepatitis, which constitutes one of the main pathways for progression to fibrosis and the development of liver cirrhosis [ 56 ]. There are numerous molecules and signaling pathways under investigation that could play a prominent role in detecting the disease at early stages and developing new therapeutic targets [ 57 , 58 ].…”

Machine learning approaches for early detection of non-alcoholic steatohepatitis based on clinical and blood parameters

“…This approach can provide the accuracy performance in range of 79–87% to diagnose NASH. Ghandian presented a comparison of the performance of different machine learning algorithms to detect the progression of NAFL to NASH 18 . He conducted a study on the electronic health record (HER) information of 700 volunteers.…”