Objective: Implement a computational system using a novel model based on neural networks and fuzzy logic to assess the risk of developing hypertension in 4 years based on previous data of patients as an alternative to the Framingham Heart Study guidelines, in this way providing physician with a tool to help them in their treatment of patients. Design and method: The artificial neural network (ANN) is trained with the risk factors of the patients, considering as the inputs: the systolic and diastolic blood pressures, age, sex, smoking, body mass index and if the patient has hypertensive parents. The ANN models the information, obtaining as output the percentage of the risk of developing hypertension in the next 4 years. Different tests are performed by varying the ANN architecture parameters, such as the number of layers, the number of neurons per layer, training algorithm, and epochs. The best results are obtained with 2 hidden layers, 10 and 14 neurons in each layer, 500 epochs and with the Levenberg-Marquart training algorithm. Results: The results obtained with the ANN model and the traditional Framingham Heart Study are compared, and it can be observed that there is no significant difference between them, that is, the neural network is learning correctly, of the set of patients used for testing, and it is observed that in only one there was a variation between the methods of 1%. Conclusions: This soft computing paradigm is powerful for these cases, as it demonstrates that it can be applied with a high level of confidence, since, as can be observed the variation between the results obtained by the formula given by the Framingham Heart Study and the artificial neural network is very low.
Read full abstract