超声颈动脉多参数智能测量方法研究

颈动脉参数测量是心血管疾病早期诊断的重要手段。近年研究揭示颈动脉多种参数的信息互补性为疾病风险评估和发病机制提供新的影响因子，提升心血管疾病的风险评估能力。结合人工智能的超声颈动脉多参数测量能显著降低手工测量所需的高昂时间和人力成本。现有研究难以解决测量输入、测量过程和测量输出三方面的不确定性带来的干扰。为此，本项目拟研究高干扰图像环境下临床多参数的人工智能推断这一关键问题，以提供可靠的血管形态学和功能学的多种参数信息。包括提出颈动脉组织图像属性和语义的关联性表达方法，以增强测量方法的稳定性；提出颈动脉图像信息与参数值的关联性表达方法，以增强智能测量方法在参数值语义的变化干扰下的有效性。提出颈动脉多参数语义的关联性表达方法，以增强智能测量方法在多参数相互影响的变化干扰下的鲁棒性。本项目的完成将实现超声颈动脉多参数测量技术，为进一步研究基于颈动脉多参数的心血管疾病诊断打下基础。

Parameter measurement of carotid artery in ultrasound images is significant in the early evaluation of cardiovascular disease (CVD). Recent studies have been focusing on information complementarity among multiple parameters. This is because that they can provide new influence factors of CVD, and thus improves the ability of CVD risk assessment. With artificial intelligence, the multi-parameter measurement of carotid artery in ultrasound images can significantly reduce the high time and labor consuming for manual parameter measurement. Existing studies are difficult to solve the bottleneck of high uncertainty in the input, the output and the procedure of the multi-parameter measurement. This project aims to study the key issue of the artificial intelligence during the multi-parameter measurement of carotid artery. This project aims to study the artificial intelligence for clinical parameter inference in the image environment with high interference. It can provide reliable multi-parameter information from the aspect of vessel morphology and function. This study firstly proposes the method to describe the relationship between the semantics of carotid parameters and the ultrasound image attributes in order to enhance the stability of the measurement method. Then, it proposes the learning strategy of image features and multiple carotid parameters to improve the effectiveness of the measurement method. Finally, it proposes the semantic relationship among multiple parameters to increase the robustness of the measurement method. The achievement of this project will realize the multi-parameter measurement of carotid artery in ultrasound images. This can facilitate the early diagnosis of CVD based on multiple carotid parameters.