A Hybrid Attention-Driven Deep Learning Model for Osteoporosis Detection in Knees

Osteoporosis is a prevalent bone disease characterized by reduced bone density and increased fracture risk, particularly among older adults and postmenopausal women. Early detection and accurate diagnosis are crucial to mitigate the morbidity and mortality associated with osteoporotic fractures. This paper proposes an advanced deep-learning methodology to enhance the accuracy and efficiency of osteoporosis detection using knee osteoporotic data. Our approach integrates features from pre-trained model, DenseNet169 and a custom designed attention model. The fused feature set from these models is then passed through a fully connected neural network to classify images as either osteoporotic or normal. The results of our proposed model demonstrated significant improvements in classification accuracy, achieving a high accuracy rate on unseen test data. Integrating DenseNet169, and attention mechanisms, effectively captured comprehensive features, resulting in robust and reliable osteoporosis detection. Moreover, the model outperformed current state-of-the-art methods for classification tasks, highlighting its potential for early intervention in clinical settings.