Automatic ICRS scoring of cartilage lesions using arthroscopic OCT images

Abstract

Articular cartilage injury is a common cause of chronic disability in both humans and animals. Current treatment strategies offer several possibilities and in order to select the optimal repair procedure, accurate determination of size and severity of a lesion is important [1,2]. Recently, an equine ex vivo study showed that arthroscopic optical coherence tomography (OCT) provides high resolution optical images of the cartilage layer [3]. Furthermore, in that study morphological characteristics of cartilage including depth of lesions could be more accurately determined than with conventional arthroscopy. Although the inter- and intra-investigator agreement by means of OCT is better than with conventional arthroscopy [43.9% versus 31.7% and 68.9% versus 56.7%, respectively] [4], reproducibility of OCT based articular cartilage grading needs improvement. Especially, avoiding the subjectivity introduced by assessment by individual surgeons would be a great step forward. In this study we aim at introducing a first version of software capable for automatic and reproducible ICRS-scoring of cartilage lesions.