This paper presents an extensive review of cardiac anomaly detection and classification in fetuses from a deep learning perspective. Anomaly detection is important because fetuses are affected by various cardiac anomalies, leading to mortality or severe complications in the later stages. Hence, early detection of anomalies is essential for further treatment. Our paper mainly discusses the application of deep learning techniques and the promising results obtained for anomaly detection in 1-Dimensional, 2-Dimensional and 3-Dimensional data.  These studies suggest that deep learning models can be effective in detecting fetal state conditions and the specific structure of the fetal heart. Some promising results include the one-dimensional convolutional Neural Network model, which achieved 97.46% accuracy in detecting the fetal state condition, and the recurrence plot Convolutional Neural Network, which achieved 98.6% accuracy in detecting fetal hypoxia.  In 2-Dimensional data, an ensemble of neural networks achieves an Area Under Curve of 0.99 for detecting cardiac substructures, and the Mask Recurrent Convolutional Neural Network provides astonishing results in multiclass detection. The cropping segmentation calibration method yielded the best results for detecting and partitioning the ventricular septum in the 3D ultrasound data. However, each method has its own merits and limitations, and future work can be conducted to improve the performance of these methods, such as using larger datasets or developing more advanced training strategies