Uzunlamasına Kan Basıncı Verilerindeki Değişimleri Tespit Etmek İçin Rassal Orman Temelli Bir İzleme Yöntemi: Retrospektif Bir Çalışma

Abstract

Objective: High blood pressure is a serious health condition. Patients and clinicians need tools to early detect deviations from normal blood pressure. Adapting modern methods of monitoring to blood pressure monitoring (BPM) has been a favorable solution. This study focuses on adaptation of machine learning to BPM. The daily chrono-biological changes effect blood pressure and may lead to false alarms in an early detection system. We provide an approach that detects changes in blood pressure with limited baseline data. Material and Methods: Our approach uses random forest as an alternative to traditional process monitoring algorithms which test each data point compared to a baseline dataset. In addition, our method converts testing problem into a supervised learning problem using a sliding baseline. We used real data and synthetic data to show the potential of the proposed method for different types of hypertension: sisto-diastolic hypertension, isolated diastolic hypertension and white coat effect. Results: Our observations support that the method can detect various patterns such as sisto-diastolic hypertension, isolated diastolic hypertension and white coat effect successfully. Conclusion: We described the development of a machine learning based monitoring approach to early detect changes in blood pressure. The proposed method (1) requires relatively small baseline data, (2) can be adapted to realtime patient data, and (3) can detect various types of hypertension.