“…Niu et al [65] proposed a neural network composed of convolutional and residual layers based on ResNet34, fine-tuned with 1832 images that contained patients with Turner syndrome and controls. They achieved 97% accuracy, a value higher than those obtained with other techniques also tested in this article: MLP trained with geometric, texture and color features, and two other deep network models: VGG16 and ResNet18.…”