Knee osteoarthritis (KOA) is a disease that grows in incidence and frequency with increasing age, to the point where in people over the age of 60, around 10% of men and 13% of women have symptomatic KOA. The illness may also affect other joints in the body. The number of persons in the age group that is associated with the most severe symptoms of KOA is continuing to rise as the population continues to age. Initially, 3D-KOA dataset is considered, which has the properties of magnetic resonance imaging of knee disorders. Then, principal component analysis (PCA) is applied, which normalized all the records in the dataset and eliminated the missing values, unknown values. Here, the PCA is also dimensionality reduction method, which extracted the statistical features from the dataset. Then, multi scale residual network (MSRN) is used to upgrade the low-resolution features into higher resolution features, which boosts the probabilities. Finally, transfer learning (TL) based AlexNet model is used to classify the normal and abnormal stages of KOA using MSRN features. The simulation results shows that the proposed model outperformed in terms of accuracy, sensitivity, specificity, as compared to existing methods.