Applying data mining techniques to medical time series: an empirical case study in electroencephalography and...

Area of Science:

• Medical Informatics
• Data Science
• Biomedical Engineering

Background:

• Medical data, particularly time series, presents challenges for knowledge extraction.
• Identifying specific events within medical time series is crucial for decision-making.
• Existing methods may not fully exploit the potential of complex medical datasets.

Purpose of the Study:

• To apply the Knowledge Discovery in Databases (KDD) process to medical time series data.
• To demonstrate the effectiveness of KDD techniques in classifying medical conditions and traits.
• To compare KDD performance against traditional neural network approaches.

Main Methods:

• Utilized patient electronic health records (EHR) containing stabilometric and electroencephalographic (EEG) time series data.
• Applied a comprehensive KDD process to analyze 396 stabilometric and 200 EEG series.
• Employed various KDD techniques for classification tasks within two distinct case studies.

Main Results:

• Achieved high classification accuracy: 99.86% and 98.11% for epilepsy diagnosis (EEG).
• Attained excellent classification accuracy: 99.4% and 99.1% for early-age sports talent (stabilometry).
• KDD techniques outperformed traditional neural network-based classification methods.

Conclusions:

• The KDD process is highly effective for extracting valuable knowledge from medical time series.
• KDD offers superior classification performance in medical domains compared to conventional methods.
• This approach holds significant promise for advancing medical decision-making and diagnostics.