Abstract:
Arthritis is a disease caused by inflammation of joints. It is the primary cause of human impairment. It affects
mostly the neck, knee, and palm of hand, elbow, lung, and heart. There are over a hundred various types of
arthritis. Osteoarthritis, rheumatoid arthritis, psoriatic arthritis, gout arthritis, and lupus arthritis are among
the most common types of arthritis. Physicians use x-ray machines to scan the damaged body of the patient,
but it is difficult to determine the types of arthritis So, Imaging processing is required for a more accurate
diagnosis of arthritis. As we've seen, earlier research works have only focused on a single type of arthritis.
Now, here we develop a Computer-aided diagnosis (CAD) knee arthritis disease classification model for the
most common occurred arthritis diseases namely Osteoarthritis, Rheumatoid arthritis, and gout arthritis. To
implement this research work, we collect 665 x-ray images from JUMC, Jimma Aweytu hospital, and
Fromsis hospital, then we apply deep learning approach CNN architecture image processing technique to
improve the accuracy of the model like image augmentation. We use image normalization like cropping size
to 100 by 100, image augmentation from 665 raw x-ray images to 1725 augmented images. Generally, we
develop a model that automatically classifies knee arthritis disease. We compare four models Vgg16,
ResNet50, DenseNet121, and our custom-developed model KneeArthritisModel by different image sizes 100
x 100, 75 x 75, and 50 x 50. We use the softmax activation function for classification, and the relu activation
function for other hidden layers. Our developed model KneeArthritisModel achieves 91% accuracy when we
use 100 x100 image size. Our model classifies the arthritis knee x-ray image into four classes of disease as
normal knee, osteoarthritis, Rheumatoid arthritis, and gout arthritis
