VALDO 在何处？MICCAI 2021 血管性病变检测与分割挑战赛
计算机视觉与模式识别
2022-08-16 v1 人工智能
摘要
脑小血管病的影像标志物提供有关脑健康的宝贵信息，但其人工评估耗时且受显著的评分者内与评分者间变异性影响。自动评级可惠及生物医学研究及临床评估，但现有算法的诊断可靠性未知。在此，我们呈现作为医学图像计算与计算机辅助干预（MICCAI）2021 国际会议卫星事件举办的血管性病变检测与分割（Where is VALDO?）挑战赛的结果。该挑战赛旨在推动自动检测与分割脑小血管病小型稀疏影像标志物的方法发展，即扩大血管周围间隙（EPVS）（任务1）、脑微出血（任务2）和假定血管源性腔隙（任务3），同时利用弱标签与噪声标签。总计 12 支队伍参与挑战赛，提出针对一项或多项的方案（任务1-EPVS 4支、任务2-微出血 9支、任务3-腔隙 6支）。训练与评估均使用多队列数据。结果显示各队伍间及各任务间性能差异巨大，任务1-EPVS 与任务2-微出血结果可喜，而任务3-腔隙尚无可实用结果。其亦凸显跨病例性能不一致性，可能阻碍个体层面使用，而在群体层面仍具效用。
引用
@article{arxiv.2208.07167,
title = {Where is VALDO? VAscular Lesions Detection and segmentatiOn challenge at MICCAI 2021},
author = {Carole H. Sudre and Kimberlin Van Wijnen and Florian Dubost and Hieab Adams and David Atkinson and Frederik Barkhof and Mahlet A. Birhanu and Esther E. Bron and Robin Camarasa and Nish Chaturvedi and Yuan Chen and Zihao Chen and Shuai Chen and Qi Dou and Tavia Evans and Ivan Ezhov and Haojun Gao and Marta Girones Sanguesa and Juan Domingo Gispert and Beatriz Gomez Anson and Alun D. Hughes and M. Arfan Ikram and Silvia Ingala and H. Rolf Jaeger and Florian Kofler and Hugo J. Kuijf and Denis Kutnar and Minho Lee and Bo Li and Luigi Lorenzini and Bjoern Menze and Jose Luis Molinuevo and Yiwei Pan and Elodie Puybareau and Rafael Rehwald and Ruisheng Su and Pengcheng Shi and Lorna Smith and Therese Tillin and Guillaume Tochon and Helene Urien and Bas H. M. van der Velden and Isabelle F. van der Velpen and Benedikt Wiestler and Frank J. Wolters and Pinar Yilmaz and Marius de Groot and Meike W. Vernooij and Marleen de Bruijne},
journal= {arXiv preprint arXiv:2208.07167},
year = {2022}
}