Intelligent Method of Non-Invasive Detection and Grading of Hypertension Using Fingertip Photoplethysmography

With the gradual increase in hypertension and the trend toward younger age in China in recent years, the existing detection methods for hypertension have a low detection rate, which can no longer meet the needs of more and more complex hypertension detection, so detection methods for hypertension needs to be supplemented. We proposed an innovative method for detecting and grading hypertension, directly obtain the details and high-dimensional features of the hypertensive patients' PPG signals through machine learning algorithms. The features of these PPG signals can reflect the damage of hypertension to the arteries. We detected and graded hypertension by the degree of arterial damage, used the fingertips PPG signals of 219 patients from Guilin People's Hospital, the selected patients have been diagnosed into three categories: normal blood pressure, prehypertension, and hypertension. By comparing the grading accuracy, sensitivity, and specificity of different models, it is shown that under the same data set and experimental platform, the model based on 14-layer deep convolutional neural network and support vector machine (Gaussian kernel, γ=0.5, C=20) is superior to other algorithms, and the accuracy of the recognition of hypertension respectively reached 92.44% and 96.92%. We provided a innovative method for grading and non-invasive detecting of hypertension.