Prediction of axillary lymph node metastasis in early breast cancer patients with ultrasonic videos based deep learning.

Objective: To develop a deep learning (DL) model for predicting axillary lymph node (ALN) metastasis using dynamic ultrasound (US) videos in breast cancer patients. Methods: A total of 271 US videos from 271 early breast cancer patients collected from Xiang'an Hospital of Xiamen University andShantou Central Hospitabetween September 2019 and June 2021 were used as the training, validation, and internal testing set (testing set A). Additionally, an independent dataset of 49 US videos from 49 patients with breast cancer, collected from Shanghai 10th Hospital of Tongji University from July 2021 to May 2022, was used as an external testing set (testing set B). All ALN metastases were confirmed using pathological examination. Three different convolutional neural networks (CNNs) with R2 + 1D, TIN, and ResNet-3D architectures were used to build the models. The performance of the US video DL models was compared with that of US static image DL models and axillary US examination performed by ultra-sonographers. The performances of the DL models and ultra-sonographers were evaluated based on accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Additionally, gradient class activation mapping (Grad-CAM) technology was also used to enhance the interpretability of the models. Results: Among the three US video DL models, TIN showed the best performance, achieving an AUC of 0.914 (95% CI: 0.843-0.985) in predicting ALN metastasis in testing set A. The model achieved an accuracy of 85.25% (52/61), with a sensitivity of 76.19% (16/21) and a specificity of 90.00% (36/40). The AUC of the US video DL model was superior to that of the US static image DL model (0.856, 95% CI: 0.753-0.959, P<0.05). The Grad-CAM technology confirmed the heatmap of the model, which highlighted important subregions of the keyframe for ultra-sonographers' review. Conclusion: A feasible and improved DL model to predict ALN metastasis from breast cancer US video images was developed. The DL model in this study with reliable interpretability would provide an early diagnostic strategy for the appropriate management of axillary in the early breast cancer patients.

Continuous transverse scanning of the fetal heart using a cross-sectional image database of common fetal congenital heart deformities.

OBJECTIVE: To provide an anatomical basis for continuous transverse scanning of the fetal heart by analyzing the typical cross-sectional characteristics of different types of congenital heart deformities (CHDs) using an anatomical image database. MATERIALS AND METHODS: The database consisted of cross-sectional images obtained from 45 cases of common fetal CHDs, which were continuously displayed by the three-dimensional software Amira 5.3.1. The following anatomical parts were observed from the database of heart samples in a bottom-to-top manner: the coronary sinus, four chambers, left ventricular outflow tract, right ventricular outflow tract, and transverse ductal and aortic arches. The anatomical characteristics of these sections were analyzed and compared with the ultrasonic transverse views obtained from the same fetuses. RESULTS: During the display of the anatomical database of 45 cases of common fetal CHDs, the aforementioned typical cross sections were successively revealed, along with the corresponding pathological features. These sections also exhibited a very good correspondence with the ultrasonic transverse views of the same cases. CONCLUSION: The database of cross-sectional anatomical images of fetal CHDs provided an anatomical basis for continuous transverse scanning of the fetal heart.

Establishment of a database of fetal congenital heart malformations and preliminary investigation of its clinical application.

OBJECTIVE: The aim of this study was to create a database of anatomical ultrathin cross-sectional images of fetal hearts with different congenital heart diseases (CHDs) and preliminarily to investigate its clinical application. MATERIALS AND METHODS: Forty Chinese fetal heart samples from induced labor due to different CHDs were cut transversely at 60-µm thickness. All thoracic organs were removed from the thoracic cavity after formalin fixation, embedded in optimum cutting temperature compound, and then frozen at -25°C for 2 hours. Subsequently, macro shots of the frozen serial sections were obtained using a digital camera in order to build a database of anatomical ultrathin cross-sectional images. RESULTS: Images in the database clearly displayed the fetal heart structures. After importing the images into three-dimensional software, the following functions could be realized: (1) based on the original database of transverse sections, databases of sagittal and coronal sections could be constructed; and (2) the original and constructed databases could be displayed continuously and dynamically, and rotated in arbitrary angles. They could also be displayed synchronically. The aforementioned functions of the database allowed for the retrieval of images and three-dimensional anatomy characteristics of the different fetal CHDs, and virtualization of fetal echocardiography findings. CONCLUSION: A database of 40 different cross-sectional fetal CHDs was established. An extensive database library of fetal CHDs, from which sonographers and students can study the anatomical features of fetal CHDs and virtualize fetal echocardiography findings via either centralized training or distance education, can be established in the future by accumulating further cases.