The prediction of hospitalization for congestive heart failure using an artificial neural network based on echocardiography

• Main Authors: Hung-Yu Yang, 楊弘宇
• Other Authors: Chien-Yeh Hsu
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/48102606556331058106

碩士 === 臺北醫學大學 === 醫學資訊研究所 === 98 === Congestive heart failure (CHF) is one of the major causes of hospitalization in population older than 65 years. Doppler echocardiography is one of the objective method to evaluate cardiac function. These evaluations let us know whether heart function is normal or impaired. The aim of this study was to design an artificial neural network (ANN) model capable of predicting the exact possibility of hospitalization due to CHF. A total 7473 Patients were included in the study from Jan. 2008 to Dec. 2008 as training cases. Another 8124 patients collected from Jan. 2009 to Dec. 2009 as test cases. Fifteen echocardiographic variables and two clinical variables were collected from hospitalization patients. ANN model was set up by training the network with data from training set and subsequently testing with data from another test set to determine the optimal ANN architecture. The optimal ANN topology was found to be a standard feed-forward, fully-connected, back-propagation multilayer perceptron. The overall accuracy rate of ANN was 97.2%, which is higher than that of logistic regression (LR) (96.6%). By using the area under the receiver operating characteristics (ROC) curve as a measure of performance, the ANN outperformed the LR (0.910±0.009 versus 0.895±0.011; p = 0.008). Therefore, the CHF hospitalization prediction model using ANN performs significantly better than using LR. But this phenomenon mandates further evaluation and research on other CHF population from other hospitals.