A non-invasive test method for type-2 diabetes mellitus by pulse waveform fitting. (July 2020)

• Record Type: Journal Article
• Authors: Ouyang, Chun; Zhou, Peng; Gan, Zhongxue
• Is Part Of: Biomedical signal processing and control. Volume 60(2020)
• Publication Date: 2020-07
• Subjects: Biomedical informatics -- Collective intelligent algorithm -- Curve fitting -- Biomedical signal processing; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2020.102000
• Languages: English
• ISSNs: 1746-8094

Abstract

Graphical abstract: Highlights: Pulse waveforms at the wrist radial artery contained information on arterial stiffness and blood viscosity. A heart-to-wrist radial artery was built by using the left ventricular model and the double-windkessel model. A collective intelligent algorithm was adopted for curve fitting to obtain a globally optimal solution. Capacitors and inductors of equivalent circuits contributed effectively to the classification of both groups. Abstract: Diabetes mellitus, a chronic disease associated with elevated accumulation of glucose in the blood, is generally diagnosed in the clinic through an invasive blood test based on several items. It is divided into two categories: type-1 and type-2 diabetes. In patients with type-2 diabetes, arterial stiffness is a common complication. In this paper, we report on a non-invasive test method for distinguishing patients with type-2 diabetes from healthy patients. Data were acquired through self-designed devices and an analysis of pulse waveforms by theoretical fitting. Pulse waveforms were collected at the radial arteries. A self-designed force sensor and its control programs were adopted for pulse data collection in the clinic, which was followed by signal amplification, filtering, and analog-digit conversion. We also established a lumped-parameters model, which consisted of the left ventricle of a heart model coupling of aortic arteries near the heart and the classical four-parameter double-windkessel model from the carotid to the radial arteries. A collective intelligent algorithm, the artificial bee colony (ABC), was selected for its fast convergency and globally optimal unique solution in curve fitting of clinic experimental results. In total, 840 cycles of pulse waveforms from 30 patients with type-2 diabetes and 52 healthy people were used in supervised learning based on the support vector machine (SVM) method. We found that the blood flow inertia factor, L, and the total compliance of peripheral arteries, C 2, from the established model, which defined the difficulty of blood flow and arterial stiffness, respectively, significantly influenced the classification of the type-2 diabetic group and the healthy group. The possible reasons were that increasing glucose and microcirculation may have changed the blood flow velocity, manifesting in L changing in the model. The established model was effective with classification rate for diabetic participants of 70.0% and enabled rapid convergency of pulse waveform fitting, indicating that it is a promising method for non-invasive testing of type-2 diabetes.