Support Vector Machines and logistic regression to predict temporal artery biopsy outcomes

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Medical Diagnostics

Background:

• Support vector machines (SVM) are advanced statistical methods.
• SVMs are suggested to be superior to traditional logistic regression for clinical classification tasks.
• Temporal artery biopsy (TABx) is a key diagnostic procedure for giant cell arteritis.

Purpose of the Study:

• To compare the predictive performance of SVM against logistic regression for temporal artery biopsy (TABx) outcomes.
• To evaluate if SVM offers an advantage in classifying patients for giant cell arteritis diagnosis.

Main Methods:

• A dataset of 530 TABx patients with 10 covariates was utilized.
• Data were randomly divided into training and testing sets.
• Area under the receiving operating curve (AUC), misclassification rate (MCR), and false negative rate (FN) were key comparison metrics. SVM was tuned using AUC and MCR.

Main Results:

• The optimal SVM model parameters were gamma = 0.01267 and cost = 26.466, with 133 support vectors.
• Logistic regression achieved an AUC of 0.827, MCR of 0.184, and FN of 0.524.
• SVM achieved an AUC of 0.825, MCR of 0.168, and FN of 0.571.

Conclusions:

• Support vector machines (SVM) did not demonstrate a significant advantage over logistic regression for predicting TABx results in this patient cohort.
• Logistic regression remains a viable and comparable method for TABx outcome prediction in giant cell arteritis.