| Autor: |
Gruss, Sascha, Velana, Maria, Werner, Philipp, de Oliveira Andrade, Adriano, Al-Hamadi, Ayoub, Walter, Steffen |
| Jazyk: |
angličtina |
| Rok vydání: |
2019 |
| Předmět: |
|
| DOI: |
10.5281/zenodo.3457608 |
| Popis: |
In the future, automatic pain monitoring may enable health care professionals to assess and manage pain objectively. In this framework, the researchers created a database of biopotentials for the development of an automatic pain-recognition system and the optimization of its performance. The goal of the current research work was to optimize pain features with a Support Vector Machine (SVM) classifier and to obtain high recognition rates for pain quantification in a two- and multiclass problem. Data of 90 participants were acquired under the induction of phasic heat pain stimulation. 13 features were finally selected from the following categories: amplitude, stationarity, linearity, variability and similarity. Classification analyses were performed with a SVM for four classification problems (baseline vs. pain threshold; baseline vs. pain tolerance; pain threshold vs. pain tolerance; baseline vs. pain threshold vs. pain tolerance). High classification accuracies were obtained for baseline vs. pain threshold (80 %) and baseline vs. pain tolerance (91 %). This study is the first to indicate one distinct set of features which successfully differentiated a no pain condition from different pain intensities. Thus, the detection of pain patterns via an automatic pain-recognition system in clinical settings is feasible more than ever before because of the small number of physiological parameters and features. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
