Diabetes is a chronic disease which can lead to Diabetic Foot Ulcer (DFU), which has been the subject of much research. In the past, many studies have used Convolution Neural Network (CNN) to analyze DFUs. However, recent research has explored Graph Neural Networks (GNNs) are more suited for analyzing data that has complex relational structures, such as the thermogram data used in DFU studies. By using GNN, both local and global information about the data were incorporated, allowing the model to make more accurate predictions. In this study, we have preprocessed the existing thermogram dataset and applied a two-layer graph convolution network. Our research compares the performance of GNN model with three machine learning models–Support Vector Machine, Logistic Regression, and Naive Bayes, over diabetic foot ulcer thermogram data for DFU classification. The Graph Neural Network model demonstrated stable and superior performance compared to the Support Vector Machine (SVM), Logistic Regression, and Naive Bayes models. The GNN model achieved an overall accuracy of 90.82%, recall of 90.71%, and an F1 score of 90.77%, along with a sensitivity of 96% and a specificity of 85.42%. Overall, the study aimed to demonstrate the potential of using GNN to analyze DFU.