跳到正文
原文
Nature 健康护理主题·· 2026-10-14AI 评分49

自适应在线产前超声图像与视频解读训练的前瞻性对照教育干预研究

Adaptive online training for prenatal ultrasound image and video interpretation: a prospective controlled educational intervention study

AI 导读

一项前瞻性对照教育干预研究评估了自适应在线学习工具在产前超声图像和视频解读训练中的效果。干预组(n=42)在完成自适应学习工具后,T2总分达82.76%,显著高于对照组(n=46)的65.83%(p<0.001,Cohen's d=1.70)。作者指出,由于对照组为非结构化自我学习且未包含非自适应对照臂,该结果不能直接证明工具优于传统结构化自学或自适应算法本身的作用。

正文

Ultrasound is the most important and commonly used imaging method for the detection of prenatal fetal pathologies, but is highly dependent on the interpreters’ knowledge. Regarding the current lack of ultrasound interpretation training in medical education, an individually adaptive online-learning tool was developed and evaluated against an unstructured control condition in which no learning activity was specified, prescribed or monitored. A prospective controlled educational intervention study with a pre–post design and quasi-random allocation to intervention and control groups was conducted. Within the learning tool, images and videos were presented together with corresponding questions to facilitate repetitive training in the recognition and assessment of relevant prenatal sonographic features. Participants in the intervention group (n = 42) completed a baseline assessment (T1) on fetal image and video interpretation before using the tool. After completing the predefined requirements of the learning tool, they underwent a comparable post-intervention assessment (T2). Participants in the control group (n = 46) completed the same T1 and T2 assessments but engaged only in unmonitored self-directed learning between the two time points. The primary outcome was the group difference in the change in overall prenatal ultrasound image and video interpretation performance from baseline assessment (T1) to final assessment (T2). Secondary outcomes included changes in image-based and video-based subscores, as well as learner satisfaction and educational benefit. Baseline performance did not differ between groups at T1 (p = .878; intervention 69,14%, control 68.78%), although baseline variance was greater in the intervention group (SD 13.11 vs. 8.50% points; Levene p = .042). At T2 the intervention group scored substantially higher (82.76% vs. 65.83%), t(86) = 7.94, p < .001 (exact p = 6.88e − 12), Cohen’s d = 1.70 (Hedges’ g = 1.68; bootstrap 95% CI 1.24 to 2.26). The group × time interaction on the overall score was F(1, 86) = 44.35, p < .001 (exact p = 2.46e − 09), partial η² = 0.340, corresponding to a between-group difference in change of 16.58% points (95% CI 11.63 to 21.52). This result survived Bonferroni correction across the full family of 15 tests, as well as permutation, rank-based and trimmed-mean tests and leave-one-out deletion of every participant. The image subscore interaction was likewise robust (F(1, 86) = 46.93, p < .001, partial η² = 0.353); the video subscore interaction was nominally significant but did not survive correction for multiplicity (p = .048, Holm-adjusted p = .335) and is reported as exploratory. Feedback was collected from the intervention group's participants.

The findings indicate that the adaptive online tool produced substantially larger gains in prenatal ultrasound image and video interpretation than an unstructured control condition over the same interval. Because the control condition was neither standardised nor monitored, and because the trial included no non-adaptive comparator, the results should not be read as evidence that the tool is superior to structured conventional self-study, nor as evidence that the adaptive algorithm specifically accounts for the effect. A three-arm trial comparing adaptive training, an equivalent non-adaptive training sequence and a no-intervention condition is required to establish both. Clinical trial registration: Not applicable. This study was a prospective controlled educational intervention study and did not constitute a clinical trial according to the WHO/ICMJE definition, as it did not prospectively assign patients or participants to a healthrelated intervention in order to evaluate biomedical or healthrelated outcomes. The intervention consisted exclusively of an educational online learning tool, and the outcomes were educational performance measures and learner feedback.

来源:Nature 健康护理主题 · nature.com