Using squeeze-and-excitation blocks to improve an accuracy of automatically grading knee osteoarthritis severity using convolutional neural networks

In this paper, we investigate the effect of squeeze-and-excitation blocks on improving classification quality osteoarthritis using convolutional neural networks ResNet and DenseNet families. We show that use these improves according to Kellgren-Lawrence scale by 1–3 % without a significant modifi-cation model structure. also demonstrate combining 0 1 classes into one class allows increase accuracy grading 12.74 %, losing information about disease. The best final ac-curacy attained was 84.66 when an ensemble three with DenseNet-121 architecture blocks, which significantly exceeds performance existing state-of-the-art. obtained results can be used both for prelimi-nary automatic diagnosis as auxiliary tool.