Machine Learning‐Based Thermal Imaging for Vulvar Intraepithelial Neoplasia Detection

Vulvar intraepithelial neoplasia (VIN) is increasing in prevalence, yet screening options remain limited and existing diagnostic methods show low accuracy. This study evaluates infrared thermal imaging as an alternative screening approach for VIN detection. We analyzed thermal images from 51 patients with histopathologically confirmed VIN, captured using a FLIR A400 thermal camera. Temperature distributions of healthy vulvar tissue were first characterized to establish baseline values. Thermal features of VIN lesions were then extracted and optimized using principal component analysis (PCA). Three machine learning models-support vector machine (SVM), random forest (RF), and linear discriminant analysis (LDA)-were trained and evaluated for VIN diagnosis. SVM demonstrated the best performance with an F1 score of 75% and accuracy of 74.19%. These findings suggest that machine learning-based infrared thermography shows promise as a non-invasive screening tool for VIN detection.