Machine learning based classification model for screening of infected patients using vital signs

Abstract

Objectives: The classification of healthy versus infected persons, and the early detection of disease sources, plays an important role in preventing spread of disease and in curing the disease. The current traditional quarantine methods using remote body thermometers as well as questionnaires have not been highly effective due environment and subjective human factors. The use of Machine Learning algorithms may be more objective and optimal for this purpose. Methods: In this paper, a non-contact measuring system using medical radar is proposed to acquire data. Then, data captured from this radar is passed through filters to both eliminate interference and to provide vital parameters such as heart and respiration rate. Finally, the classification between healthy and infected people is executed by using five Machine learning algorithms. With the measured dataset, the classification models are built through training and test steps. Results: The classification results of the algorithms are evaluated based on the f1-score parameter with accuracy greater than 80%. In particular, the Deep Learning algorithms gives the highest result of 98%. Conclusion: This study implements patient classification algorithms, which achieved good performance. This might be beneficial for rapid screening of infected patients at public health centers in underdeveloped areas, where people have little access to healthcare. Motivation & significance: The classification of healthy and infected people can help prevent the spread of disease in a community. With such relatively accurate results, in the future, the system can be directly applied in practice.

Keywords

• Dengue fever
• Medical radar
• Non-contact measurement
• Machine learning