Parkinson's disease (PD) is a common neurological disorder that can severely affect the patient's quality of life. The Archimedean spiral drawing test is a widely used diagnostic tool for PD detection. Given the advancements in computer vision techniques, deep learning approaches have been widely adopted for detecting PD from spiral drawing images, achieving promising performance. However, they often lack the explainability required in clinical settings. In this study, we designed an explainable machine learning approach to support PD diagnosis from spiral drawing images, using a set of features that are clinically relevant. The trained model is able to produce competitive results, and the explanations of prediction results reveal that the model tends to focus on critical diagnostic factors and produce predictions that correspond to clinical findings. This approach helps to align model predictions with clinical understanding and promote greater trust in AI among users.