Most research on Parkinson’s Disease detection using handwriting relies on the online approach (i.e., digital tablets and smart pens) rather than the offline one (i.e., scanned paper documents), primarily because it captures dynamic temporal features that reveal early motor symptoms. However, offline handwriting presents some notable advantages, such as: its lower cost, fewer technical barriers especially for older adults who more frequently suffer from this disease, access to more historical data from patients, and easier clinical integration, among others. This paper describes a study to discriminate handwriting from healthy and Parkinsonian patients using offline handwriting images reconstructed from the online data contained in the PaHaW dataset. For this purpose, we perform a collection of experiments that have produced F1-Score prediction results above 62% for some writing tasks.