Anam Mustaqeem1, Syed Muhammad Anwar2, Abdul Rashid Khan3, Muhammad Majid4. 1. Department of Software Engineering, University of Engineering and Technology, Taxila, Pakistan. 2. Department of Software Engineering, University of Engineering and Technology, Taxila, Pakistan. Electronic address: s.anwar@uettaxila.edu.pk. 3. Cardiology Department, Pakistan Ordinance Factories Hospital, Wah, Pakistan; Wah Medical College, Wah, Pakistan. 4. Department of Computer Engineering, University of Engineering and Technology Taxila, Pakistan.
Abstract
OBJECTIVES: An intelligent information technology based system could have a positive impact on the life-style of patients suffering from chronic diseases by providing useful health recommendations. In this paper, we have proposed a hybrid model that provides disease prediction and medical recommendations to cardiac patients. The first part aims at implementing a prediction model, that can identify the disease of a patient and classify it into one of the four output classes i.e., non-cardiac chest pain, silent ischemia, angina, and myocardial infarction. Following the disease prediction, the second part of the model provides general medical recommendations to patients. METHODS: The recommendations are generated by assessing the severity of clinical features of patients, estimating the risk associated with clinical features and disease, and calculating the probability of occurrence of disease. The purpose of this model is to build an intelligent and adaptive recommender system for heart disease patients. The experiments for the proposed recommender system are conducted on a clinical data set collected and labelled in consultation with medical experts from a known hospital. RESULTS: The performance of the proposed prediction model is evaluated using accuracy and kappa statistics as evaluation measures. The medical recommendations are generated based on information collected from a knowledge base created with the help of physicians. The results of the recommendation model are evaluated using confusion matrix and gives an accuracy of 97.8%. CONCLUSION: The proposed system exhibits good prediction and recommendation accuracies and promises to be a useful contribution in the field of e-health and medical informatics.
OBJECTIVES: An intelligent information technology based system could have a positive impact on the life-style of patients suffering from chronic diseases by providing useful health recommendations. In this paper, we have proposed a hybrid model that provides disease prediction and medical recommendations to cardiac patients. The first part aims at implementing a prediction model, that can identify the disease of a patient and classify it into one of the four output classes i.e., non-cardiac chest pain, silent ischemia, angina, and myocardial infarction. Following the disease prediction, the second part of the model provides general medical recommendations to patients. METHODS: The recommendations are generated by assessing the severity of clinical features of patients, estimating the risk associated with clinical features and disease, and calculating the probability of occurrence of disease. The purpose of this model is to build an intelligent and adaptive recommender system for heart diseasepatients. The experiments for the proposed recommender system are conducted on a clinical data set collected and labelled in consultation with medical experts from a known hospital. RESULTS: The performance of the proposed prediction model is evaluated using accuracy and kappa statistics as evaluation measures. The medical recommendations are generated based on information collected from a knowledge base created with the help of physicians. The results of the recommendation model are evaluated using confusion matrix and gives an accuracy of 97.8%. CONCLUSION: The proposed system exhibits good prediction and recommendation accuracies and promises to be a useful contribution in the field of e-health and medical informatics.
Authors: Robin De Croon; Leen Van Houdt; Nyi Nyi Htun; Gregor Štiglic; Vero Vanden Abeele; Katrien Verbert Journal: J Med Internet Res Date: 2021-06-29 Impact factor: 5.428