KneeOsteo.WRN-50-2

Knee Osteoarthritis (KOA) is the most common types of Osteoarthritis (OA) and it is diagnosed by physicians using the widely used 0–4 Kellgren Lawrence (KL) grading system which places the KOA in a spectrum of 5 grades; starting from normal (0) to Severe OA (4). In this paper, we propose a transfer learning approach of a very deep residual learning based network (Wide ResNet-50-2) which is fine-tuned using X-ray plain radiographs from the Osteoarthritis Initiative (OAI) dataset to learn the KL severity grading of KOA. In addition, we propose a data augmentation approach of OAI data to avoid imbalance and reduce overfitting by applying it only to certain KL grades depending on their number of plain radiographs. Experimentally, we tested the model on original and augmented x-rays and it showed good generalization power in predicting the KL grade of knee x-rays with an accuracy of 72% and F1-Score of 74%. Moreover, using Grad-Cam, we also showed that network is selecting some correct features when predicting a KL grade of a knee radiograph. Such achieved results showed that our such a model outperformed several other related works, and it can be further improved to be used to help radiologists make more accurate and precise diagnosis. In case you use this work please cite our paper as the following: Helwan, Abdulkader, Azar, Danielle, and Abdellatef, Hamdan. ‘An Update on the Knee Osteoarthritis Severity Grading Using Wide Residual Learning’. 1 Jan. 2022 : 1 – 13.

https://pubmed.ncbi.nlm.nih.gov/35848003/