Application of Deep Neural Network for Calculation of Pretest Probability of the Heart Diseases
Presentation Type
Event
Abstract
Application of deep learning techniques for calculation of pre-test probability(PTP) scoring of heart diseases based on the structured and unstructured data reduces unnecessary cardiac imaging tests, medical costs, and other potential risks. For any patient, the calculation of PTP will increase the performance of a given diagnostic test for heart diseases and be performing no test for a patient with fewer than a certain threshold of the likelihood function. In the model for cardiac imaging decision-making system, the input layer is the features of the patients’ medical history and the different physical and biological attributes. The output layer is defined as the decision based on the probability for every heart diseases tests and the several combinations of them. The Deep Neural Network (DNN) model whose network uses the efficient Adam gradient descent optimization algorithm with a logarithmic loss function, was implemented using Keras. The number of neurons, number of hidden layers, batch size, convergence criterion, activation function, learning rate and regularization parameters is defined as the model requirement. The pretest probability of the heart diseases is compared relatively against the score from traditional linear measures for the model validation.
Start Date
2-12-2018 12:00 PM
Application of Deep Neural Network for Calculation of Pretest Probability of the Heart Diseases
University Student Union: Volstorff A
Application of deep learning techniques for calculation of pre-test probability(PTP) scoring of heart diseases based on the structured and unstructured data reduces unnecessary cardiac imaging tests, medical costs, and other potential risks. For any patient, the calculation of PTP will increase the performance of a given diagnostic test for heart diseases and be performing no test for a patient with fewer than a certain threshold of the likelihood function. In the model for cardiac imaging decision-making system, the input layer is the features of the patients’ medical history and the different physical and biological attributes. The output layer is defined as the decision based on the probability for every heart diseases tests and the several combinations of them. The Deep Neural Network (DNN) model whose network uses the efficient Adam gradient descent optimization algorithm with a logarithmic loss function, was implemented using Keras. The number of neurons, number of hidden layers, batch size, convergence criterion, activation function, learning rate and regularization parameters is defined as the model requirement. The pretest probability of the heart diseases is compared relatively against the score from traditional linear measures for the model validation.