A novel method for dipper/non-dipper pattern classification in hypertensive and non-diabetic patients.

Background: In the classical process, it was proven that ABPM data were the most significant attributes both by physician and ranking algorithms for dipper/non-dipper pattern classification as mentioned in our previous papers. To explore if any algorithm exists that would let the physician skip this diagnosis step is the main motivation of the study.

Objective: The main goal of the study is to build up a classification model that could reach a high-performance metrics by excluding ABPM data in hypertensive and non-diabetic patients.

Methods: The data used in this research have been drawn from 29 hypertensive patients without diabetes in endocrinology clinic of Marmara University in 2011. Five of 29 patient data were later removed from the dataset because of null data.

Results: The findings showed that dipper/non-dipper pattern can be classified by artificial neural network algorithms, the highest achieved performance metrics are accuracy 87.5%, sensitivity 71%, and specificity 94%.

Conclusions: This novel method uses just two attributes: Ewing-score and HRREP. It offers a fast and low-cost solution when compared with the current diagnosis procedure. This attribute reduction method could be beneficial for different diseases using a big dataset.