Abstarct: Abstract Bone age estimation is one of the most important challenges in pediatrics. It assists physicians in examining and monitoring the growth phases of a child up to puberty and even beyond until the growth plates fuse. The importance of this issue becomes apparent when physicians can earlier detect developmental disorders such as short stature, abnormal bone growth, or sudden increase in height and take necessary actions. Since deep learning has made remarkable progress over the past few decades, numerous methods have been proposed to automate this process. Previous studies have generally focused on processing raw images or manual preprocessing, which has led to increased error or lack of generalizability. Another weakness of previous methods is their slow speed in training and evaluation. However, preprocessing is considered one of the most important components of bone age estimation, as it significantly aids deep neural networks in estimating bone age compared to processing raw images. The method proposed in this thesis automates the process of estimating bone age baxsed on the patient's gender and hand X-ray image, and announces the result with higher accuracy and speed compared to previous studies. The error rate in this process has been reduced to 3.81 months in the age range of 1 to 228 months on the Radiological Society of North America (RSNA) Pediatric Bone Age test dataset containing 1400 data. The processing time is about 1.64 seconds per data, allowing physicians to use this method concurrently.