The relationship between HbA1c and ultrasound plaque textures in atherosclerotic patients. Issue 1 (December 2016)

• Record Type: Journal Article
• Authors: Huang, Xiao-Wei; Zhang, Yan-Ling; Meng, Long; Qian, Ming; Zhou, Wei; Zheng, Rong-Qin; Zheng, Hai-Rong; Niu, Li–Li
• Is Part Of: Cardiovascular diabetology. Volume 15:Issue 1(2016)
• Page Start: 1
• Page End: 8
• Publication Date: 2016-12
• Subjects: Ultrasound imaging -- Carotid plaque -- Texture -- HbA1c -- Cardiovascular disease; Diabetes -- Complications -- Periodicals; Cardiovascular system -- Diseases -- Complications -- Periodicals; Cardiovascular system -- Diseases -- Prevention -- Periodicals; 616.462
• DOI: 10.1186/s12933-016-0422-5
• Languages: English
• ISSNs: 1475-2840

Abstract

Abstract Objective Diabetes mellitus (DM) is associated to the morphological and componential characteristics of atheromatous plaques. It has proven that plaque textures are related to plaque components and beneficial for atherosclerotic risk stratification. The aim of this study is to compare plaque textures in patients with and without DM, and examine the relationship between HbA1c levels and the ultrasound plaque textures in atherosclerotic patients. Methods A total of 136 participants (among them 66 are diabetic and 70 are non-diabetic) suffering from carotid plaques were included. About 300 texture features were extracted from the ultrasound images of plaques using the algorithms of histogram, absolute gradient, run-length matrix, gray-level co-occurrence matrix, autoregressive model and wavelet transform, respectively. Thirty optimal features were selected by the Fisher coefficient and the mutual information measure. The most discriminating feature (MDF) was obtained from the linear discriminant analysis for the optimal features. Linear regression model was performed to investigate the relationship between HbA1c and MDF. The receiver operating characteristics (ROC) curve was further developed to validate the relation between the estimated HbA1c (models output) and diabetes status. Results A total of 12 texture features showed statistical difference between patients with and without DM. The MDF was significant higher in non-diabetic patients (0.326 ± 0.049) than diabetic patients (−0.346 ± 0.052) (p < 0.001). The optimal regression model (r = 0.348, p < 0.001) for HbA1c included a constant (p < 0.001) and the MDF (p < 0.001). The areas under ROC curve used to estimate HbA1c was 0.828. Conclusions The results indicate that there is a quantitative relationship between the HbA1c levels and plaque textures in ultrasonic images of atherosclerotic patients, which may suggest that texture analysis of the ultrasonic image of plaque is a promising method for evaluating the cardiovascular risk caused by DM in patients with plaques.