Chatbot to predict need of a stent in cardiac treatment

Abstract

Background: Every year, across the globe about 1,800,000 people suffers with heart attack and India has the highest rate in the world with almost 25% families having a member who suffered heart attack at least once. One of the solutions to reduce the heart attacks in patients is placing a stent. Once the blockage is identified, it is suggested that a stent with the required size is inserted into the coronary artery. The invasive device and its associated cost automatically leads to a dilemma for both the doctors as well as patients. The young practitioners of cardiology may face a dilemma sometimes whether to go for stent or not. The patients undergoing treatment for cardiac issues would be in a two state mind whether to go for stent or not. The above two dilemmas can be mitigated and confidence of both doctors and patients in decision making can be enhanced with the help of a technology driven application that suggests requirement of stent with a degree of confidence.  Methods: The chatbot in our project shall help in taking textual and image based inputs of the information of the heart and coronary artery of a patient and predicts the requirement of stent being placed. The real time data of 50 patients was used in training the model with a data set consisting of 349 scanned images of angiogram. The model developed consists of sequential CNN model with 5 convolutional 2D layers and 5 Maxpool layers.  Results: The model could achieve an accuracy of 81%. Conclusions: The authors have successfully developed and tested a machine learning based chatbot model to predict the need of a stent in cardiac treatment.