Accelerometer-derived classifiers for early detection of degenerative joint disease in cats.

Abstract

Decreased mobility is a clinical sign of degenerative joint disease (DJD) in cats, which is highly prevalent, with 61 % of cats aged six years or older showing radiographic evidence of DJD. Radiographs can reveal morphological changes and assess joint degeneration, but they cannot determine the extent of pain experienced by cats. Additionally, there is no universal objective assessment method for DJD-associated pain in cats. Developing an accurate evaluation model could enable earlier treatment, slow disease progression, and improve cats' well-being. This study aimed to predict early signs of DJD in cats using accelerometers and machine learning techniques. Cats were restricted to indoors or limited outdoor access, including being walked on a lead or allowed into enclosed areas for short periods. Fifty-six cats were fitted with collar-mounted sensors that collected accelerometry data over 14 days, with data from 51 cats included in the analysis. Cat owners assessed their cats' mobility and assigned condition scores, validated through clinical orthopaedic examinations. The study group comprised 24 healthy cats (no owner-reported mobility changes) and 27 unhealthy cats (owner-reported mobility changes, suggestive of early DJD). Data were segmented into 60-second windows centred around peaks of high activity. Using a Support Vector Machine (SVM) algorithm, the model achieved 78 % (confidence interval: 0.65, 0.88) area under the curve (AUC), with 68 % sensitivity (0.64, 0.77) at 75 % specificity (0.68, 0.79). These results demonstrate the potential of accelerometry and machine learning to aid early DJD diagnosis and improve management, offering significant advances in non-invasive diagnostic techniques for cats.