Scholarlink

<p>Computer vision is a rapidly growing field with tools emerging to support clinical practice. This systematic review and critical appraisal synthesizes the applications of computer vision in vascular surgery. MEDLINE, Embase, Web of Science and Cochrane CENTRAL were searched from inception to June 28, 2024. Vascular diseases studied, data sources, methods and outcomes were recorded. Critical appraisal was conducted using the PROBAST and TRIPOD+AI guidelines. Overall, 288 studies were included with an exponential rise from 2017 onwards. The majority of studies addressed aortic pathologies (33%), carotid stenosis (30%), and foot ulcers (25%), while few focused on peripheral artery disease (6%). Most were observational using retrospective (81%) or prospective data (15%), and one clinical trial was included. Dice coefficient (51%) and accuracy (36%) were the most commonly reported performance metrics with infrequent use of AUROC (17%). Only 15% of studies had a low risk of bias and overall adherence to the TRIPOD+AI checklist was relatively poor at 57%. Overall, we suggest greater attention to peripheral artery disease; for dice coefficient and AUROC to be used in segmentation and discrimination tasks, respectively; and for the TRIPOD+AI statement to be consulted early on during model development.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Computer vision applications in vascular surgery: a systematic review and critical appraisal

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月18日

查看全文
  • Annudesh Liyanage,
  • Ben Li,
  • Jason Yi,
  • Muhammad Mamdani,
  • Konrad Salata

摘要

Computer vision is a rapidly growing field with tools emerging to support clinical practice. This systematic review and critical appraisal synthesizes the applications of computer vision in vascular surgery. MEDLINE, Embase, Web of Science and Cochrane CENTRAL were searched from inception to June 28, 2024. Vascular diseases studied, data sources, methods and outcomes were recorded. Critical appraisal was conducted using the PROBAST and TRIPOD+AI guidelines. Overall, 288 studies were included with an exponential rise from 2017 onwards. The majority of studies addressed aortic pathologies (33%), carotid stenosis (30%), and foot ulcers (25%), while few focused on peripheral artery disease (6%). Most were observational using retrospective (81%) or prospective data (15%), and one clinical trial was included. Dice coefficient (51%) and accuracy (36%) were the most commonly reported performance metrics with infrequent use of AUROC (17%). Only 15% of studies had a low risk of bias and overall adherence to the TRIPOD+AI checklist was relatively poor at 57%. Overall, we suggest greater attention to peripheral artery disease; for dice coefficient and AUROC to be used in segmentation and discrimination tasks, respectively; and for the TRIPOD+AI statement to be consulted early on during model development.