背景:我们的目标是确定教育游戏SonoQz是否可以提高卵巢肿瘤超声评估的诊断性能。
方法:开发了SonoQz移动应用程序,作为医生进行超声评估的教育工具,基于卵巢肿瘤的静止图像。游戏包含来自324个卵巢肿瘤的图像,手术前由超声专家检查。训练阶段,参与者在SonoQz应用程序中评估了至少200个案例,之前是训练前测试,然后是训练后的测试。两个相等的测试(A和B),每个由20个案例组成,被用作训练前和训练后的测试。一半的用户先测试了A,B秒,其余的以相反的顺序进行了测试。用户被要求根据国际卵巢肿瘤分析(IOTA)简单规则对肿瘤进行分类(1),(2)良性或恶性,和(3)提示特定的组织学诊断。具有固定效应的Logistic混合模型,用于训练前和训练后的测试,并交叉了参与者和案例的随机效应,用于确定考试成绩的任何改善,灵敏度,和特异性。
结果:来自19个医疗中心的58名医生参加。比较训练前和训练后的测试,正确分类病例的中位数,在简单规则评估中从72%增加到83%,p<0.001;在将病变分类为良性或恶性肿瘤时,从86%到95%,p<0.001;在从43%到63%的特异性诊断中,p<0.001。当肿瘤分类为良性或恶性时,在灵敏度水平不变的情况下(98%与97%,p=0.157),特异性从70%增加到89%,p<0.001。
结论:我们的结果表明,教育游戏SonoQz是一种有效的工具,可以提高评估卵巢肿瘤的诊断能力,特别是通过减少假阳性的数量,同时保持高灵敏度。
BACKGROUND: Our objective was to determine whether the educational game SonoQz can improve diagnostic performance in ultrasound assessment of ovarian tumors.
METHODS: The SonoQz mobile application was developed as an educational tool for medical doctors to practice ultrasound assessment, based on still images of ovarian tumors. The game comprises images from 324 ovarian tumors, examined by an ultrasound expert prior to surgery. A training phase, where the participants assessed at least 200 cases in the SonoQz app, was preceded by a pretraining test, and followed by a posttraining test. Two equal tests (A and B), each consisting of 20 cases, were used as pre- and posttraining tests. Half the users took test A first, B second, and the remaining took the tests in the opposite order. Users were asked to classify the tumors (1) according to International Ovarian Tumor Analysis (IOTA) Simple Rules, (2) as benign or malignant, and (3) suggest a specific histological diagnosis. Logistic mixed models with fixed effects for pre- and posttraining tests, and crossed random effects for participants and cases, were used to determine any improvement in test scores, sensitivity, and specificity.
RESULTS: Fifty-eight doctors from 19 medical centers participated. Comparing the pre- and posttraining test, the median of correctly classified cases, in Simple Rules assessment increased from 72% to 83%, p < 0.001; in classifying the lesion as benign or malignant tumors from 86% to 95%, p < 0.001; and in making a specific diagnosis from 43% to 63%, p < 0.001. When classifying tumors as benign or malignant, at an unchanged level of sensitivity (98% vs. 97%, p = 0.157), the specificity increased from 70% to 89%, p < 0.001.
CONCLUSIONS: Our results indicate that the educational game SonoQz is an effective tool that may improve diagnostic performance in assessing ovarian tumors, specifically by reducing the number of false positives while maintaining high sensitivity.