Diagnosis of chronic liver disease from liver scintiscans by fuzzy reasoning

We investigated the possibility that fuzzy reasoning might be used to standardize diagnosis of liver disease based on scintigraphic results and compared the results with those obtained when scintiscans were scored conventionally.

Seventy-five patients with chronic liver disease (11 patients had chronic persistent hepatitis, 26 had chronic aggressive hepatitis and 38 had cirrhosis) and 25 controls were studied. Another 75 patients with hepatitis or cirrhosis were examined to test the effectiveness of the membership functions. Liver scintiscans were taken 20 min after the intravenous injection of 111 MBq of 99mTc-phytate. Fuzzy reasoning was used to evaluate the following five items: the ratio of the sizes of the left and right lobes, splenomegaly, radioactivity in the bone marrow, deformity of the liver and distribution of radioactivity in the liver. The degree of conformity to each of the three liver diseases being investigated was substituted into the membership function for the conclusion. The center of gravity for each patient's results was calculated. Conventional scoring was made with three levels for each of the five items examined by fuzzy reasoning.

Distinctions between chronic persistent hepatitis and chronic aggressive hepatitis were difficult to assess with fuzzy reasoning and conventional scoring. The diagnostic accuracy was 95% for patients with cirrhosis and 88% for patients with chronic hepatitis with fuzzy reasoning. With conventional scoring the accuracy was 86% for patients with cirrhosis and 75% for patients with chronic hepatitis. When fuzzy reasoning was used to examine the other 75 patients with chronic liver diseases, the accuracy was 93% for patients with cirrhosis and 86% for patients with chronic hepatitis.

The method is simple and can be used routinely in clinical settings.