Background and aims
Temperature asymmetry between the affected and contralateral extremity is common in patients with CRPS (Complex Regional Pain Syndrome). In the acute phase of CRPS the affected extremity is generally warmer than the contralateral extremity. In the chronic phase, the affected extremity can become cold due to vasomotor disturbances that cause decreased blood flow.  

Although clinically temperature differences are mostly  not difficult to estimate, in research a more objective method for assessing temperature differences is needed, desirable expressed in one value to make statistical analyses possible.
Previous studies have shown that skin-temperature has a high correlation with blood flow. 

One way to measure skin temperature is through videothermography. In this study we investigate whether radiomics can be used to objectively quantify and classify thermal images. 

Methods
Thermal images are made of both patients with unilateral CRPS and age and gender-matched control subjects. All images are first semi-automatically segmented, followed by extraction of histogram and texture features. Feature selection is performed using principal component analysis. Lastly, support vector machines are used to develop a classification model. The performance is estimated through cross-validation.

Results
This study is ongoing. Currently, more images are made and machine learning is used to train the first classifiers. We will show our first results.

Conclusions
The ultimate goal is to develop a model that detects and quantifies thermal asymmetry. An objective and valid method can contribute to a more reliable diagnosis and evaluation of the effects of treatment of vasomotor dysregulation.

OBJECTIVE ASSESSMENT OF VIDEO THERMOGRAPHY IMAGES OF PATIENTS WITH COMPLEX REGIONAL PAIN SYNDROME