 This paper proposes using convolutional neural networks, CNNs, to assist clinicians in determining the severity of dental crowding and the necessity of tooth extraction for orthodontic treatment planning. Three CNN models, ResNet 50, ResNet 101, and VGG 19, were evaluated on a dataset of 3136 orthodontic occlusal photographs with annotations by two orthodontists. The results show that the maxillary and mandibular VGG 19 models had the lowest mean error of 0.84 mm and 1.06 mm, respectively, for teeth landmark detection. Additionally, the maxillary VGG 19 model had the highest accuracy, 0.922, an area under the curve, AUC 0.961, for tooth extraction diagnosis. These findings suggest that CNNs can assist clinicians in the diagnosis and decision-making of treatment plans. This article was authored by Jiho Ryu, Yee Heung Kim, Tai Wu Kim, and others. We are article.tv, links in the description below.