Machine Learning-Based Thermal Imaging for Vulvar Intraepithelial Neoplasia Detection

Area of Science:

• Medical imaging
• Oncology
• Machine learning

Background:

• Vulvar intraepithelial neoplasia (VIN) prevalence is rising.
• Current VIN screening methods have limitations and low diagnostic accuracy.

Purpose of the Study:

• To evaluate infrared thermal imaging as a novel screening tool for VIN detection.
• To assess the accuracy of machine learning models in diagnosing VIN using thermal imaging data.

Main Methods:

• Collected thermal images from 51 patients with confirmed VIN using a FLIR A400 thermal camera.
• Established baseline temperature distributions for healthy vulvar tissue.
• Extracted and optimized thermal features using principal component analysis (PCA).
• Trained and evaluated Support Vector Machine (SVM), Random Forest (RF), and Linear Discriminant Analysis (LDA) models for VIN diagnosis.

Main Results:

• Support Vector Machine (SVM) model achieved the highest performance.
• SVM model demonstrated an F1 score of 75% and an accuracy of 74.19% for VIN diagnosis.

Conclusions:

• Machine learning-based infrared thermography is a promising non-invasive screening approach for VIN detection.
• This technology could improve early detection and management of VIN.