Pubic Symphysis-Fetal Head Segmentation and Angle of Progression

The high risk of maternal and perinatal morbidity is associated with longer labor duration due to the slow progression of fetal descent, but accurate assessment of fetal descent by monitoring the fetal head (FH) station remains a clinical challenge in guiding obstetric management. Based on clinical findings, the transvaginal digital examination is the most commonly used clinical estimation method of fetal station. However, this traditional approach is very subjective, often difficult, and unreliable. The need of an objective diagnosis found its solution in the use of transperineal ultrasound (TPU) able to assess FH station by measuring the angle of progression (AoP) that is the extension the FH goes through in its descent. Manual segmentation of symphysis pubis (SP)-fetal head from ITU images for clinical radiologists is considered as the most reliable but extremely time-consuming procedure prone to subjectivity and large inter-observer variability. With the rapid development of artificial intelligence in medical images, automatic measurement algorithms based on ITU images are expected to solve the above problems. This segmentation competition requires the use of MHA files for image data, which includes 4000 samples of both original images and ground truth labels. The original images have a shape format of 3x256x256. The ground truth labels have a shape format of 256x256, and contain pixels labeled as 0, 1, or 2, where 0 represents the background, 1 represents the pubic symphysis, and 2 represents the fetal head. For the competition evaluation, only prediction images with a size of 256x256 are accepted. The prediction images can only contain pixels labeled as 0, 1, or 2, with 0 representing the background, 1 representing the pubic symphysis, and 2 representing the fetal head. _________________________________________________________________________________________________________________________________________________________________________________ PSFHS is the test dataset. Each needs to send a email including name and affiliations to the organizer via email: [e-mail]产妇及围产期并发症的高风险与胎儿下降缓慢导致的产程延长密切相关，但通过监测胎头（fetal head, FH）位置来准确评估胎儿下降情况，仍是产科临床诊疗决策中的一大挑战。基于临床实践，经阴道指检是目前临床最常用的胎头位置评估方法，但该传统方法主观性极强、操作难度大且结果不可靠。针对客观诊断的需求，经会阴超声（transperineal ultrasound, TPU）提供了解决方案：它可通过测量产程进展角（angle of progression, AoP）——即胎头下降过程中所经过的角度——来评估胎头位置。 临床放射科医师对经会阴超声图像中的耻骨联合（symphysis pubis, SP）与胎头进行手动分割，被认为是最可靠的标注方式，但该流程极为耗时，且易受主观性影响，存在观察者间差异较大的问题。随着人工智能在医学影像领域的快速发展，基于经会阴超声图像的自动测量算法有望解决上述痛点。 本次分割竞赛要求使用MHA格式文件作为图像数据，共包含4000组原始图像与真值标签样本。原始图像的尺寸格式为3×256×256；真值标签的尺寸格式为256×256，像素仅包含0、1、2三种取值，其中0代表背景，1代表耻骨联合，2代表胎头。 竞赛评审仅接收尺寸为256×256的预测结果图像，且预测图像的像素仅可包含0、1、2三种取值，对应规则与真值标签一致：0为背景、1为耻骨联合、2为胎头。 PSFHS为本次竞赛的测试数据集。参赛者需发送包含姓名及所属机构信息的邮件至主办方邮箱：[e-mail]。