Determining the condition of lower limb Varus and Valgus deformity is one of the initial steps in osteotomy and arthroplasty procedures. To ascertain this state, a series of angles is often measured manually, with the quality of the measurement greatly relying on the experience of the individual taking it. There have been many attempts to predict the weight-bearing line (WBL) ratio using simple knee radiographs. This paper suggests a convolutional neural network-supported method for determining the necessary angles HKA in lower limb radiography (X-ray) images by detecting ROI as an initial step for accuracy on radiograph and moreover reduce effort and time of radiologist. To obtain high accuracy in the measurement process, a training dataset based on geometric information linked to the deformity correction principles is constructed in addition to using a decentralized deep learning method that includes two orders for the radiography. The performance of the proposed method is contrasted with standard references made up of manually. The proposed methodology involves the use of the reptile search algorithm for optimizing model hyper parameters and the evaluation of a neural network model using various performance indicators and visual aids to assess its performance in categorizing joint coordinates. The AUC score of 0.9839 indicates strong discriminatory capacity in properly categorizing positive and negative occurrences.