ARVNet: Attention-Residual V-Net for Fetal Cardiac Rhabdomyoma Segmentation in Echocardiography

Fetal cardiac rhabdomyoma, while infrequent, is a serious entity that can affect fetal development and well-being. Early detection with the help of sophisticated imaging methods, like echocardiography, is essential for controlling and avoiding complications. This paper examines the application of deep learning for the automation of the task of segmenting echocardiographic images for identifying fetal cardiac rhabdomyomas. Using an ARVNet deep learning model, the objective is to develop a model that can accurately detect and segment tumor areas in fetal heart ultrasounds. The model is trained using a dataset of labeled echocardiographic images with methods such as data augmentation and preprocessing for enhanced performance. The segmentation task is aimed at detecting tumor areas relative to normal tissue with high accuracy. Measuring criteria like precision, recall, and F1-score are applied to evaluate the efficiency of the model in the identification of rhabdomyomas, particularly keeping false negatives low, which is important for early diagnosis. This approach serves to improve clinicians' diagnostic strengths through a faster and more consistent means of fetal cardiac tumor detection, ultimately serving towards making the best decisions for treatment and management. The suggested method may prove to be a useful instrument in fetal heart imaging, providing an important step toward more effective and automated healthcare solutions for pregnant women.