Gastroenterology, University of Kansas School of Medicine, Kansas City, Kansas, United States.
Gastroenterology, University of Colorado Anschutz Medical Campus, Aurora, Colorado, United States.
Endoscopy. 2017 Dec;49(12):1243-1250. doi: 10.1055/s-0043-116015. Epub 2017 Aug 14.
Experts can accurately predict diminutive polyp histology, but the ideal method to train nonexperts is not known. The aim of the study was to compare accuracy in diminutive polyp histology characterization using narrow-band imaging (NBI) between participants undergoing classroom didactic training vs. computer-based self-learning. Trainees at two institutions were randomized to classroom didactic training or computer-based self-learning. In didactic training, experienced endoscopists reviewed a presentation on NBI patterns for adenomatous and hyperplastic polyps and 40 NBI videos, along with interactive discussion. The self-learning group reviewed the same presentation of 40 teaching videos independently, without interactive discussion. A total of 40 testing videos of diminutive polyps under NBI were then evaluated by both groups. Performance characteristics were calculated by comparing predicted and actual histology. Fisher's exact test was used and < 0.05 was considered significant. A total of 17 trainees participated (8 didactic training and 9 self-learning). A larger proportion of polyps were diagnosed with high confidence in the classroom group (66.5 % vs. 50.8 %; < 0.01), although sensitivity (86.9 % vs. 95.0 %) and accuracy (85.7 % vs. 93.9 %) of high-confidence predictions were higher in the self-learning group. However, there was no difference in overall accuracy of histology characterization (83.4 % vs. 87.2 %; = 0.19). Similar results were noted when comparing sensitivity and specificity between the groups. The self-learning group showed results on a par with or, for high-confidence predictions, even slightly superior to classroom didactic training for predicting diminutive polyp histology. This approach can help in widespread training and clinical implementation of real-time polyp histology characterization.
专家可以准确预测微小息肉的组织学特征，但培训非专业人员的理想方法尚不清楚。本研究旨在比较接受课堂理论培训与基于计算机的自学的参与者在使用窄带成像（NBI）对微小息肉组织学特征进行描述时的准确性。
两所机构的学员被随机分为课堂理论培训组或基于计算机的自学组。在理论培训中，经验丰富的内镜医生查看了关于 NBI 模式下腺瘤性和增生性息肉以及 40 个 NBI 视频的演示，并进行了互动讨论。自学组则独立查看了相同的 40 个教学视频，没有互动讨论。然后，两组学员都要对 40 个在 NBI 下的微小息肉的测试视频进行评估。通过比较预测和实际的组织学结果来计算性能特征。使用 Fisher 确切检验， < 0.05 被认为具有统计学意义。
共有 17 名学员参与（8 名课堂理论培训，9 名基于计算机的自学）。课堂组对更大比例的息肉做出了高度置信的诊断（66.5%比 50.8%； < 0.01），尽管在高度置信预测中，自学组的敏感性（86.9%比 95.0%）和准确性（85.7%比 93.9%）更高。然而，组织学特征的总体准确性没有差异（83.4%比 87.2%； = 0.19）。当比较两组之间的敏感性和特异性时，也得到了类似的结果。自学组在预测微小息肉组织学方面的结果与课堂理论培训相当，对于高度置信的预测甚至略优于课堂理论培训。这种方法有助于广泛开展实时息肉组织学特征培训和临床应用。
