OrthoX-AI: Automated X-Ray Alignment and Correction System [ Ilizarov Technique ]

Manual assessment of lower limb alignment using Lower Limb X-rays is a time-consuming and expertise-driven process, critical in orthopedic diagnosis and deformity correction. We propose OrthoX-AI, a fully automated deep learning pipeline for Evaluating lower limb malalignment using anatomical angle prediction. Our system integrates a CNN-based keypoint detection model to localize hip, knee, and ankle joints, followed by an angle regression model predicting eight clinically relevant angles: mLPFA, aMPFA, mLDFA, aLDFA, mMPTA, aMPTA, mLDTA, and aLDTA. The predicted angles are analyzed via a CORA-based rule engine to classify the limb as normal, varus, or valgus. Evaluated on a self-curated dataset of 2000+ annotated X-rays, our method achieves a mean absolute error (MAE) of \(\sim \) 5 \(^\circ \) in angle prediction and over 90% classification accuracy. OrthoX-AI provides a deployable, ONNX-exportable clinical tool for orthopedic imaging.