Advancements in medical imaging and computational techniques have propelled the development of automated systems for medical image processing, like segmentation, detection, classification, and measurements from the images. Automating the measurement of the Femur-Tibia Angle (FTA) and Medial Proximal Tibial Angle (MPTA) from X-rays offers benefits over manual methods in large healthcare systems. In radio-graphs, anatomical variability among individual subjects, differences in the nature of knee deformity, and the presence of implants lead to difficulty in automatically detecting landmarks like the knee center, the tibial, and the femoral axis required for measuring the angles mentioned above. Also, the presence and the size of the implants may vary from subject to subject in post-operative images, making it challenging to develop a universal solution to automate feature detection and segmentation. This work envisages developing an automated computer workflow to locate the landmarks like knee center, tibial, and femoral axis to measure FTA, MPTA, and
Hip Knee Ankle Angle (HKAA) from pre- and post-operative X-rays. As part of this work, we propose a deep convolution neural network model, vSegNet, to segment the region between the femur and tibia and compare its performance for the chosen application with the existing state-of-the-art neural network models. In addition, we evaluated the performance of the proposed model, vSegNet, on public datasets. Application of the encoder part of the proposed neural network vSegNet, in classifying the spinal metastases is also demonstrated.