The Parkinson disease diagnosis can be challenging as there is no definitive test for it, and the likelihood of misdiagnosis is particularly high when performed by non-specialists. One clinical assessment used to evaluate fine motor skills, hand-eye coordination, and tremors in individuals with neurological disorders, including Parkinson’s, is represented by the spiral and wave drawing tests. These tests involve asking a person to trace or draw a spiral pattern and a wave pattern on paper or a digital tablet. The smoothness and steadiness of the hand movements are measured, and any irregularities or tremors can provide critical information for healthcare professionals in diagnosing or monitoring Parkinson’s disease, essential tremors, or other movement disorders. In this paper, we present a method to automatically analyze spiral and wave drawing tests to determine if a subject has Parkinson’s disease. We utilize three different Convolutional Neural Networks i.e., DenseNet, MobileNet, and EfficientNet, achieving an accuracy of 0.95 on a dataset of 4683 spiral drawing tests and an accuracy equal to 0.92 in the evaluation of a dataset composed of 4686 wave drawing tests, demonstrating the method’s effectiveness. Additionally, to offer insight into the model’s predictions, our method can visualize the areas of the spiral and wave drawing test that the model associates with Parkinson’s disease by exploiting two different Class Activation Mapping algorithms.