Detail publikace

Advanced Parkinson's Disease Dysgraphia Analysis Based on Fractional Derivatives of Online Handwriting

Originální název

Advanced Parkinson's Disease Dysgraphia Analysis Based on Fractional Derivatives of Online Handwriting

Anglický název

Advanced Parkinson's Disease Dysgraphia Analysis Based on Fractional Derivatives of Online Handwriting

Jazyk

en

Originální abstrakt

Parkinson's disease (PD) is one of the most frequent neurodegenerative disorder with progressive decline in several motor and non-motor skills. Due to time-consuming and partially subjective conventional PD diagnosis, several more effective approaches based on signal processing and machine learning, e.\,g. online handwriting analysis, have been proposed. This paper introduces a~new methodology of PD dysgraphia analysis based on fractional derivatives applied in PD handwriting quantification. The proposed methodology was evaluated on a~database that consists 33~PD patients and 36~healthy controls who performed several handwriting tasks. Employing random forests classifier in combination with 5 kinematic features based on fractional-order derivatives we reached 90\,\% classification accuracy, 89\,\% sensitivity, and 91\,\% specificity. In comparison with the results of other related works dealing with the same database, the proposed approach brings improvements in PD dysgraphia diagnosis and confirms the impact of fractional derivatives in kinematic analysis.

Anglický abstrakt

Parkinson's disease (PD) is one of the most frequent neurodegenerative disorder with progressive decline in several motor and non-motor skills. Due to time-consuming and partially subjective conventional PD diagnosis, several more effective approaches based on signal processing and machine learning, e.\,g. online handwriting analysis, have been proposed. This paper introduces a~new methodology of PD dysgraphia analysis based on fractional derivatives applied in PD handwriting quantification. The proposed methodology was evaluated on a~database that consists 33~PD patients and 36~healthy controls who performed several handwriting tasks. Employing random forests classifier in combination with 5 kinematic features based on fractional-order derivatives we reached 90\,\% classification accuracy, 89\,\% sensitivity, and 91\,\% specificity. In comparison with the results of other related works dealing with the same database, the proposed approach brings improvements in PD dysgraphia diagnosis and confirms the impact of fractional derivatives in kinematic analysis.

BibTex


@inproceedings{BUT149482,
  author="Ján {Mucha} and Jiří {Mekyska} and Marcos {Faúndez Zanuy} and Karmele {Lopez-de-Ipina} and Vojtěch {Zvončák} and Zoltán {Galáž} and Tomáš {Kiska} and Zdeněk {Smékal} and Luboš {Brabenec} and Irena {Rektorová}",
  title="Advanced Parkinson's Disease Dysgraphia Analysis Based on Fractional Derivatives of Online Handwriting",
  annote="Parkinson's disease (PD) is one of the most frequent neurodegenerative disorder with progressive decline in several motor and non-motor skills. Due to time-consuming and partially subjective conventional PD diagnosis, several more effective approaches based on signal processing and machine learning, e.\,g. online handwriting analysis, have been proposed. This paper introduces a~new methodology of PD dysgraphia analysis based on fractional derivatives applied in PD handwriting quantification. The proposed methodology was evaluated on a~database that consists 33~PD patients and 36~healthy controls who performed several handwriting tasks. Employing random forests classifier in combination with 5 kinematic features based on fractional-order derivatives we reached 90\,\% classification accuracy, 89\,\% sensitivity, and 91\,\% specificity. In comparison with the results of other related works dealing with the same database, the proposed approach brings improvements in PD dysgraphia diagnosis and confirms the impact of fractional derivatives in kinematic analysis.",
  booktitle="10th International Congress on Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT)",
  chapter="149482",
  doi="10.1109/ICUMT.2018.8631265",
  howpublished="print",
  year="2018",
  month="november",
  pages="77--82",
  type="conference paper"
}