García-Mejido José Antonio, Galán-Paez Juan, Solis-Martín David, Martín-Morán Marina, Borrero-Gonzalez Carlota, Fernández-Gomez Alfonso, Fernández-Palacín Fernando, Sainz-Bueno José Antonio
Department of Surgery, Faculty of Medicine, University of Seville, Seville, Spain.
Department of Computer Science and Artificial Intelligence, Faculty of Mathematics, University of Seville, Seville, Spain.
J Clin Ultrasound. 2025 Sep;53(7):1470-1478. doi: 10.1002/jcu.24050. Epub 2025 Apr 25.
To create and validate a machine learning(ML) model that allows for identifying the correct capture of the midsagittal plane in a dynamic ultrasound study, as well as establishing its concordance with a senior explorer and a junior explorer.
Observational and prospective study with 90 patients without pelvic floor pathology. Each patient was given an ultrasound video where the midsagittal plane of the pelvic floor was recorded at rest and during the Valsalva maneuver. A segmentation model was used that was trained on a previously published article, generating the segmentations of the 90 new videos to create the model. The algorithm selected to build the model in this project was XGBoost(Gradient Boosting). To obtain a tabular dataset on which to train the model, feature engineering was carried out on the raw segmentation data. The concordance of the model, of a junior examiner and a senior examiner, with the expert examiner was studied using the kappa index.
The first 60 videos were used to train the model and the last 30 videos were reserved for the test set. The model presented a kappa index 0.930(p < 0.001) with very good agreement for detection of the correct midsagittal plane. The junior explorer presented a very good agreement (kappa index = 0.930(p < 0.001)). The senior explorer presented a kappa index 0.789(p < 0.001) (good agreement) for detection of the correct midsagittal plane.
We have developed a model that allows determining the correct midsagittal plane captured through dynamic transperineal ultrasound with a level of agreement comparable to or greater than that of a junior or senior examiner, using expert examiner assessment as the gold standard.
创建并验证一个机器学习（ML）模型，该模型能够在动态超声研究中识别正中矢状面的正确捕捉情况，并确定其与资深检查者和初级检查者的一致性。
对90例无盆底病变的患者进行观察性前瞻性研究。为每位患者提供一段超声视频，记录盆底在静息状态和瓦尔萨尔瓦动作时的正中矢状面。使用在先前发表的一篇文章中训练的分割模型，对90个新视频进行分割以创建模型。本项目中选择用于构建模型的算法是XGBoost（梯度提升）。为了获得用于训练模型的表格数据集，对原始分割数据进行了特征工程处理。使用kappa指数研究模型、初级检查者和资深检查者与专家检查者的一致性。
前60个视频用于训练模型，后30个视频留作测试集。该模型的kappa指数为0.930（p < 0.001），在检测正确的正中矢状面方面具有非常好的一致性。初级检查者具有非常好的一致性（kappa指数 = 0.930（p < 0.001））。资深检查者在检测正确的正中矢状面方面的kappa指数为0.789（p < 0.001）（一致性良好）。
我们开发了一个模型，该模型能够通过动态经会阴超声确定正确捕捉的正中矢状面，其一致性水平与初级或资深检查者相当或更高，以专家检查者的评估作为金标准。
