This research proposes a deep learning-based early diagnostic model for two significant bone diseases, osteoporosis, and arthritis, which can be diagnosed from MRI images. Both these diseases affect millions of elderly people around the world and become highly dangerous if diagnosed late. The reason is that the alteration in bone density is relatively minimal in osteoporosis, and, in fact, traditional methods of diagnosis will often miss the case of osteoporosis at an early stage of the disease. We applied convolutional neural network (CNN) architectures specifically, ResNet50, VGG19, VGG16, and Inception V3 to a dataset of MRI images preprocessed for optimal feature extraction. The dataset comprises 2,800 images, with 70% used for training and the remaining 30% for testing. Among the models, ResNet50 showed better performance with an accuracy of 97.86%, followed by VGG19 at 95.6%, which indicates its effectiveness in the early identification of bone disease. Positive confirmation of the diagnostic reliability of the model can be seen in key performance metrics such as accuracy, sensitivity, specificity, and F1 score. Confirmation that deep learning models are very effective in the detection of subtle patterns within MRI scans makes this research precious for the early identification of diseases. When applied in real-time clinical settings, this model may assist radiologists in automating preliminary diagnosis, which will expedite quicker and more accurate assessments and, in turn, enable early intervention. It also has potential use in telemedicine, where it will provide high diagnostic accuracy, especially for remote or underserved areas, hence enhancing access to quality healthcare and supporting preventive healthcare strategies.