Evaluating Radiomics Feature Reduction for Thyroid Nodule Segmentation in Thermal Imaging Publisher



Etehadtavakol M^{1, 2} ; Etehadtavakol M^{1, 2} ; Moallem G⁵ ; Ng EYK⁶
Source: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Published:2025

Abstract

Radiomics encompasses a variety of image attributes, including intensity, texture, shape, and spatial relationships among pixels. In this study, we applied four feature selection techniques: Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), t-Distributed Stochastic Neighbor Embedding (t-SNE), and Recursive Feature Elimination (RFE), to reduce the dimensionality of radiomic features extracted from thyroid thermal images. Subsequently, we employed three models: U-Net with a VGG16 backbone, U-Net with ResNet50 backbone, and U-Net with DenseNet121 backbone to segment thyroid nodules. The U-Net with a VGG16 backbone combined with the LDA model achieved an average Dice coefficient of 0.4035, significantly outperforming the other models. These results highlight the potential of feature reduction techniques in enhancing thyroid nodule segmentation in thermal imaging. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.