Bone age assessment helps to detect and schedule treatment for several disorders. Estimating bone age differs from determining physical maturity based on the individual's date of birth. Assessing bone age determines development and progress, identifying and treating juvenile illnesses. Challenges in bone age assessment notably arise from poor-quality X-images, obscured bone structures, and the complexity of feature extraction due to degraded image quality, significantly impacting model performance. The proposed methodology involves utilizing pre-trained neural networks—InceptionV3, DenseNet201, XceptionNet, and MobileNetV2—finetuned by adding dense layers alongside dropout and kernel initializers adjustments. The hyperparameters for each pre-trained model are rigorously defined, and the performance evaluation encompasses a spectrum of optimizers such as Adam, Nadam, Adamax, RMSprop, and SGD. Notably, the implementation of the Adamax optimizer yields superior results, demonstrating exceptional accuracy in bone age assessment on the RSNA dataset, particularly with DenseNet201, InceptionV3, and XceptionNet models. This research presents a comprehensive comparative analysis showcasing the enhanced accuracy of bone age estimation.