Hepatitis B compensated stage liver cirrhosis screening model building method based on random forest algorithm

Abstract

The invention relates to a hepatitis B compensated stage liver cirrhosis screening model building method based on a random forest algorithm. The method comprises the steps of collecting data; preprocessing data; building a hepatitis B compensated stage liver cirrhosis screening classification model based on the random forest algorithm; testing and evaluating the classification model. According to the technical scheme, serum indexes, type-B ultrasonic imaging features and other sample data, collected in recent three years, of hepatitis B patients and compensated stage liver cirrhosis patients are used for building a noninvasive hepatitis B compensated stage liver cirrhosis forewarning model. On the basis of the model, whether a hepatitis B patient (to-be-tested sample) is in the period of compensated stage liver cirrhosis or not can be effectively predicted, and the disease progression and prognosis conditions of the hepatitis B patient can be evaluated. On the basis of evaluation of the model, unnecessary liver tissue biopsy pathological examinations can be reduced.

Description

technical field [0001] The invention relates to an evaluation and screening model establishment method for compensated liver cirrhosis, in particular to a method for establishing a hepatitis B compensated liver cirrhosis screening model based on a random forest algorithm. Background technique [0002] Liver cirrhosis is a common wasting system disease in clinical practice, and its pathogenesis is mainly related to viral hepatitis, long-term drinking, long-term medication, long-term diarrhea and family history. For patients with advanced liver cirrhosis, the clinical features are more typical, so the diagnosis is easier. However, most patients with advanced liver cirrhosis are in the decompensated stage, and some patients have even developed liver failure. Therefore, it is particularly important to be screened out in the compensated stage. If liver cirrhosis is still in the compensated stage and can be diagnosed and treated promptly, the fibrosis of the liver tissue is expec...

AI Technical Summary

Problems solved by technology

[0017] However, as far as the existing random forest algorithm is concerned, since the Gini index is used to select split nodes to construct the decision tree, the constructed decision tree is sensitive to the unbalanced class distribution, which leads to an overall decline in the random forest classification performance.

The main reason is that when the class distribution in the data set is unbalanced, the degree of confounding of the samples contained in the data set will be reduced, especially when the class distribution is highly skewed, the degree of confounding will be further reduced

At this time, when the Gini index is used to select the split attribute, the reduction in confusion (gain) may be small or even negatively increased, so that the split attribute cannot be effectively selected, and eventually the decision tree stops growing prematurely, and the minority class samples cannot be effectively divided.

[0018] On the other hand, great progress has been made in non-invasive screening methods for liver cirrhosis at home and abroad, and with the development of new medical technologies, this progress will be further expanded; however, these non-invasive screening models are mostly Single-category indicators and the model is mostly a linear combination of selected indicators, ignoring the fact that there are often nonlinear relationships in practice

Images