Fig 1.
Machine learning analysis flow chart.
Fig 2.
Accuracy, sensitivity, and specificity of ReHo (regional homogeneity)-based classification.
Table 1.
Demographic and clinical characteristics of cirrhotic patients.
Fig 3.
Discriminative ReHo-map for minimal hepatic encephalopathy (MHE) subjects.
The map is composed of 4,000 features as the optimized classification performance is achieved. The color intensity indicates the attribute weight of a feature in support vector machine (SVM) classification. The map shown includes clusters with > 50 voxels. The positive and negative weights indicate relatively increased and decreased ReHo values, respectively, in MHE patients as compared to NHE group.
Fig 4.
Correlation between test margin (i.e. the distance from SVM optimal hyperplane) and diagnostic criteria.
The blue and red circles indicate the subjects without and with MHE, respectively.
Table 2.
Most important regions discriminating between MHE and NHE subjects.
Fig 5.
Permutation distribution of the estimate (repetitions: 10,000) as the 4,000 most discriminating features were used in the linear support vector machine classifier.
GR0 is the generalization rate obtained by the classifier trained on the real class labels. With the generalization rate as the statistic, the classifier learned the relationship between the data and labels with a probability of being wrong < 0.0005.
Fig 6.
Correlation map of ReHo value and Psychometric Hepatic Encephalopathy Score (PHES) in the MHE group.
Significant positive correlations were found in the anterior cingulate gyrus and medial prefrontal cortex.