Predictive model of synovial membrane degradation using semi-automated morphometry and artificial neural networks

Gonarthrosis is a degenerative disease that affects mainly older people, but whose incidence has increased significantly in the last decade in population under the age of 65. The main objective of this study was developing a predictive model of synovial membrane degradation in relation to local nerve structures in patients with knee osteoarthritis, based on advanced morphometry and artificial neural networks (ANNs). We present here a pilot test of the method, describing preliminary findings in analyzing a pre-set number of images. We tested the system on a pre-defined set of 50 images from patients suffering of gonarthrosis in different stages. Biological material used for the histological study was synovial membrane fragments. We included 50 anonymized images from 25 consecutive patients. We found significant differences between mean fractal dimensions (FDs) of histological elements of normal and pathological tissues. In the case of immunohistochemistry, we found statistically relevant differences for mean FDs of all antibodies. We fed the data to the ANN system designed to recognize pathological regions of the examined tissue. We believe that further study will have an important contribution to the development and will bring new local targeted therapies. These could slow or reverse joint damage and pain relief in patients with osteoarthritis.