Skip to main content
Fig. 3 | Arthritis Research & Therapy

Fig. 3

From: High-throughput quantitative histology in systemic sclerosis skin disease using computer vision

Fig. 3

Prediction of SSc status in a secondary cohort. A) We developed and independently tested a logistic regression model composed of Quantitative Image Features (QIFs) to predict SSc vs. healthy control. The model output was a Diagnostic Score, i.e. the predicted probability that the image was from a patient with SSc. Box plots of the of the cross-validation (CV) and independently tested Diagnostic Scores show significant separation between groups. Using a hard threshold of 0.5, the model achieved a 1.9% misclassification rate using cross validation in the training data and 6.6% on the independent testing data set. B) As an alternative visualization of the model performance, the ROC curves for the logistic regression model show that the model achieves high area under the curve (AUC) for both CV (AUC =1.00) and testing (AUC =0.99)

Back to article page
\