FOCAL LIVER-DISEASE - NEURAL NETWORK-AIDED DIAGNOSIS BASED ON CLINICAL AND LABORATORY DATA

Objectives. - The purpose of this study was to evaluate the rise of cl inical and laboratory data in the diagnosis of benign or malignant foc al liver disease. Methods. - Diagnosis was made by artificial neural n etwork (NN), a system of simple computing units connected in a specifi c structural network. Seven clinical and laboratory variables were ret rospectively studied in 172 patients with a liver mass 193 benign, 79 malignant) detected with ultrasound. The diagnostic efficacy of NN was compared with a score based on the logistic regression model (Beaujon score). Results. - Although the sensitivity of the Beaujon score and the neural network was similar (4 malignant tumors inversely classifie d), neural network-aided diagnosis was characterized by higher specifi city and accuracy (respectively 98.9% vs 82.5%, P < 0.001, and 97.1% v s 88.4%, P < 0.002). Conclusion. - In patients with a hepatic mass, ne ural network is a valuable method for differentiating malignant and be nign tumors.