Several congenital aortic pathologies, such as Marfan syndrome and aortic coarctation, require imaging for diagnosis, although these methods have limitations in assessing complete blood flow dynamics. Computational fluid dynamics (CFD) offers a noninvasive alternative by creating virtual models from medical images. We propose the combination of convolutional neural networks and CFD as an alternative tool to classify aortic flow images and thus detect these conditions early. No previous study has applied CFD with convolutional neural networks for this purpose. In this study, we used six convolutional neural networks pretrained on 452 images (388 pathological and 64 healthy). MobileNetV2 and ResNet-18 achieved the highest percentages but were more computationally demanding, while the Compact network achieved excellent accuracy with lower computational costs and a faster inference time per image, making it suitable for real-time applications.