Radiographic knee osteoarthritis detection using liquid neural networks

Knee osteoarthritis (KOA) is a non-traumatic, slowly progressing form of arthritis that affects millions globally. The radiographic assessment of the knee can only be effective if conducted as early as possible, but routine radiographic investigation strategies frequently battle with exactness and consistency. Human diagnosis is subjective based on experience, whereas artificial intelligence (AI) techniques have shown promise, surpassing human diagnosis in some cases. In this research, the field of radiographic knee osteoarthritis detection using AI is being progressed by utilizing Liquid Neural Networks (LNN). CNN, LSTM, cross-breed models combining CNN and LSTM, CNN with bidirectional LSTM (bi-LSTM), and LNN are tested. The models were prepared on a comprehensive radiographic dataset of osteoarthritic knee joints. The LNN model accomplished a classification accuracy of 91.57%, illustrating efficiency in recognizing osteoarthritis-related irregularities. This research highlights that liquid neural networks are an effective tool for radiographic knee osteoarthritis classification.