Estimating the fibrosis stage in chronic hepatits C patients using image processing methods on ultrasonographic images – preliminary results
Aim: When diagnosing chronic hepatitis, usual ultrasonography is not safe enough when determining the difference between certain fibrosis stages. We tried to assess the usefulness of the computerized texture image analysis in noninvasive assessment of fibrosis stage in chronic hepatitis C patients.
Methods: We've selected 58 chronic hepatitis C patients, which have pure fibrosis (F0-F4) without any steatosis. We used a Logiq 7 scanner with the same working parameters for all patients. On each evaluated image, we established a Region of Interest and we extracted 166 features using 4 algorithms from it. The values resulted after applying the algorithms were analyzed and sorted according to their statistical relevance for each fibrosis stage (we used t-Student's test for a p<0.05)
Results: From the 4 tested algorithms the GLCM statistics gave us the most relevant features in order to differentiate fibrosis grades. Using all the generated features we observed that in differentiating F0 from F1 9 features (18.36%) were statistically relevant, 13 features (26.53%) – in differentiating F1-F2, 47 features (95.9%) – in differentiating F2-F3 and 32 features (65.3%) – in differentiating F3-F4.
Conclusions: Although there is no a single feature that has a relevance over 95% in all comparison cases, our findings suggest that a combination of features must be used in order to successfully diagnose the fibrosis stage. The direct medical benefit will resume in a possible distinction between patients without significant fibrosis and those with severe fibrosis or cirrhosis, distinction which is a very important prognosis factor.