Related Experiment Videos
[Study on decision support system for the interpretation of laboratory data by an artificial neural network--with a
Y Okamoto1, H Nakano, M Yoshikawa
1Department of Clinico-Laboratory Diagnostics, Nara Medical University, Kashihara.
Abstract:
An artificial neural network (ANN) was trained on laboratory data of liver function tests to estimate for histological diagnosis in patients with liver diseases such as chronic hepatitis, liver cirrhosis and fatty liver. ANN diagnosed accurately 95.3% of the cases for training and 50% of the test cases. Elimination of ZTT, GOT or A/G ratio from the data set for input reduced the ability of accurate diagnosis, whereas elimination of TBA raised this ability to 66.7%. On the other hand, accuracy of physician in diagnosing the test cases ranged from 20.8 to 62.5%. From our results it is expected that ANN may support a physician's decision on the interpretation of laboratory data.