Mimicking the immune system to diagnose Parkinson's disease from handwriting

• Published in: International Conference on Pattern Recognition pp. 2496 - 2502
• Main Authors: Parziale, Antonio, Cioppa, Antonio Della, Marcelli, Angelo
• Format: Conference Proceeding
• Language: English
• Published: IEEE 21.08.2022
• Subjects: Behavioral sciences; Detectors; Feature extraction; Parkinson's disease; Sociology; Support vector machines; Training
• ISSN: 2831-7475
• DOI: 10.1109/ICPR56361.2022.9956516

We introduce a method adopting the Negative Selection Algorithm, which mimics the way the human immune system learns to discriminate body cells from external antigens, for the computer-aided diagnosis of Parkinson's disease from online handwriting. The major advantage of the proposed method with respect to the current state-of-the-art machine learning methods is that it is trained only on data from healthy subjects, thus avoiding the burden of collecting patients' data. Moreover, it has only two parameters to set, and its implementation is by far simpler than those of most of, if not all, the methods proposed in the literature. The performance of the proposed method is evaluated on the PaHaW dataset, which includes handwriting samples drawn by 75 subjects. The results show that it outperforms the state-of-the-art methods and uses fewer features.