Application of Artificial Neural Network Model Established by Tumor Markers and Bronchofibroscopic Data in Auxiliary Diagnosis of Lung Cancer

Abstract

Abstract: Objective To establish the artificial neural network (ANN) model of auxiliary diagnosis of lung cancer combined with tumor markers and picture data collected by bronchofibroscope. Methods The levels of serum Carcinoembryonic antigen (CEA), Neuron specific enolase (NSE), Squamous cell carcinoma antigen (SCC-Ag) and Cytokeratin 19 fragment (CYFRA21-1) were detected by enzyme linked immunosorbent assay (ELISA) in 55 lung cancer patients and 64 patients with lung benign disease. The bronchofibroscopic picture characteristics were selected and quantificated, then 3 ANN intellectual models were developed, which were model only with tumor markers, only with bronchofibroscopic data, and both with them. Results Using the 3 ANN models to distinguish lung cancer in samples, the results of ANN model established by combined data were the best: its sensitivity, specificity and accurate rate were 94.5%, 96.9%, and 95.8%, respectively. Conclusion ANN model combined with tumor markers and bronchofibroscopic data can be used as a potential useful tool in auxiliary diagnosis of lung cancer.