Background:
Gout and calcium pyrophosphate dihydrate deposition disease (CPPD) are common, cause significant morbidity, and differ in their treatment. Diagnosis is usually confirmed through synovial fluid analysis, in particular through the application of polarised light microscopy. It has been recognised that current diagnostic methods lack sensitivity and specificity, which can in turn lead to under- and over-treatment and thereby excess morbidity. The role of novel computer imaging techniques to aid the interpretation of light microscopy has not been explored.
Methods:
Approximately 50 samples of synovial fluid with known diagnosis (via conventional assessment) will be analysed in a deidentified manner. Each sample will be used to generate approximately 20 images of distinct high-powered fields under polarised light microscopy, with the polariser set at randomly determined angles.
Images from half of the samples will be used to generate a training set of images. Deep learning computer imaging methods will be applied to this training set to generate an image analysis algorithm. This algorithm will then categorise images obtained from the remaining samples.
Results:
As data collection is ongoing, preliminary data and a description of the image processing algorithm will be presented.
Conclusions:
Computer imaging analysis could plausibly be applied in the future to automate the interpretation of polarised light microscopy. This holds promise for improving the specificity and sensitivity of synovial fluid analysis in crystal arthropathies.