Novel algorithm for non-invasive assessment of fibrosis in NAFLD

Abstract

Introduction: Various conditions of liver disease and the downsides of liver biopsy call for a non-invasive option to assess liver fibrosis. A non-invasive score would be especially useful to identify patients with slow advancing fibrotic processes, as in Non-Alcoholic Fatty Liver Disease (NAFLD), which should undergo histological examination for fibrosis.

Patients/methods: Classic liver serum parameters, hyaluronic acid (HA) and cell death markers of 126 patients undergoing bariatric surgery for morbid obesity were analyzed by machine learning techniques (logistic regression, k-nearest neighbors, linear support vector machines, rule-based systems, decision trees and random forest (RF)). Specificity, sensitivity and accuracy of the evaluated datasets to predict fibrosis were assessed.

Results: None of the single parameters (ALT, AST, M30, M60, HA) did differ significantly between patients with a fibrosis score 1 or 2. However, combining these parameters using RFs reached 79% accuracy in fibrosis prediction with a sensitivity of more than 60% and specificity of 77%. Moreover, RFs identified the cell death markers M30 and M65 as more important for the decision than the classic liver parameters.

Conclusion: On the basis of serum parameters the generation of a fibrosis scoring system seems feasible, even when only marginally fibrotic tissue is available. Prospective evaluation of novel markers, i.e. cell death parameters, should be performed to identify an optimal set of fibrosis predictors.

Figure 1. Prediction performance.

The ROC curve of the RF is shown (p = 0.008). Black dot: performance of the 3 nn (p = 0.02); white circle: performance of the DT (p value = 0.099). The dashed line marks the performance by chance.

Figure 2. Boxplots of the outputs of the RFs.

On the y-axis the predicted class probabilities for stage 1 and stage 2 are plotted. Generally, the RFs give higher prediction values for subjects from stage 1 compared to subjects from stage 2. The upper and the lower quartiles, the median (bold line) as well as outliers (circles) are shown.

Figure 3. Decision Tree.

The decision tree (DT) focuses on the parameters M30, M65 and AST (GOT), which is in partial agreement with the RFs. At the first level, the DT focuses on M30. If the M30 value for a given sample is less than 472.7, it is assigned to the negative class (0). If the M30 value is greater than 472.7, it is transferred to the next level that uses M65. If the M65 value for the given sample is greater than 462.1, it is assigned to the negative class, otherwise it is transferred to the last level. The last level in the DT focuses on GOT. If the GOT value for the given sample is less than 27.5, the sample is assigned as negative, otherwise as a positive sample (1).