
Bayesian prediction of liver fibrosis stage: Combinatorial use of multiple elastography and serum fibrosis markers
How much confidence does one have for determining the fibrosis stage from an elastography result without conducting a biopsy? If the value obtained through elastography is very high, the presence of cirrhosis is almost certain. In addition, a sufficiently low value can confirm a “no fibrosis” condition. However, in cases where neither a very high or a low value is obtained through elastography, the condition is always difficult to diagnose.
Bayesian prediction or Bayesian inference is a method of inference wherein, using Bayes’ rule, the probability estimate for a hypothesis is updated as additional evidence is acquired. For the staging of liver fibrosis using various elastography and serum fibrosis markers, the probability of each stage of fibrosis is predicted by updating the information of these fibrosis predictors.
Only the mean value and standard deviation from elastography or serum fibrosis markers for each fibrosis stage—i.e., F0 to F4—are required for Bayesian prediction. If the empirical values cannot be obtained, the probability estimate from MR elastography (MR touch, GE Healthcare) and ultrasound transient elastography (Fibroscan, Echosens) for each fibrosis stage can also be estimated, using the default value reported by the University of Yamanashi, Japan.
In order to obtain the required results, the measured value for the patient should be inserted into the box labeled “observed value.” After clicking “Calculate,” the Bayesian probabilities of each fibrosis stage estimated from the elastography values are obtained. Thus, we can obtain the estimated probability of each fibrosis stage for each elastography value as well as the probability of their combination.
Therefore, instead of comparing between methods for the optimal option, we should apply them in combination in order to enhance patient care using noninvasive methods.
Department of Radiology
University of Yamanashi
Yamanashi, Japan
Motosugi U, et al. Bayesian prediction for liver fibrosis staging: Combined us
e of elastography and serum fibrosis markers. Hepatology. 2012 Nov 21. doi: 10.1002/hep.26144. [Epub ahead of print]





No need to divide F1 from F0? 



