Osteoporosis is a silent bone disease that decreases bone density and puts one at risk for fractures. Traditional diagnosis using DEXA scans is Expensive , Not always accessible and Detects disease at later stages Osteoporosis is a progressive disease of bones with decreased Bone Mineral Density and increased risk of fractures, which often goes undiagnosed till serious complications develop. This study presents a new idea of using multiple modalities of machine learning to predict early osteoporosis by combining clinical data, lifestyle factors and bone X-ray imaging. Unlike the traditional methods, which use a single modality of data, the proposed method uses a hybrid architecture that incorporates deep learning and conventional machine learning techniques. Convolutional Neural Networks (CNNs), such as ResNet, are used to extract discriminative features from X-ray images and gradient boosting models are used for processing structured clinical data such as age, BMI, calcium levels and indicators of vitamin D deficiency. A novel attention-based feature fusion mechanism is proposed to effectively fuse heterogeneous data sources in order to improve the prediction accuracy. Additionally, techniques to explain artificial intelligence (XAI) such as SHAP are integrated to obtain interpretability and clarity in decision-making to facilitate clinical adoption. Experimental results show that the proposed model achieves an accuracy of more than 94% that is significantly better than baseline models. The system is further optimized for lightweight deployment to allow for its use in healthcare settings with low resources. This is an approach that provides a cost-effective, scalable, and interpretable solution to detect osteoporosis in early stages and identify risks.