Visual Interpretation of Biomedical Time Series Using Parzen Window-Based Density-Amplitude Domain Transformation

Area of Science:

• Biomedical signal processing
• Data visualization
• Machine learning for healthcare

Background:

• Biomedical time series analysis is crucial for diagnostics.
• Current methods often struggle with complex signal patterns.
• Visual analysis of raw or frequency-domain data can be limited.

Purpose of the Study:

• To propose a new method for visual analysis of biomedical time series.
• To evaluate the effectiveness of the density-amplitude domain for signal classification.
• To compare the proposed method against traditional time-amplitude and frequency-amplitude domains.

Main Methods:

• Utilized two public biomedical datasets.
• Computed density coefficients using the Parzen Windowing method.
• Visualized and classified data in the density-amplitude domain using SVM, KNN, and Naïve Bayes classifiers.
• Compared results with raw time-amplitude and frequency-amplitude representations.

Main Results:

• Visual interpretation in the density-amplitude domain yielded better classification than raw data.
• The density-amplitude representation improved classification success by up to 55% over time-amplitude methods.
• Classification success increased by up to 75% compared to frequency-amplitude methods.

Conclusions:

• The density-amplitude domain offers a superior approach for biomedical time series analysis and classification.
• This method enhances visual interpretability and diagnostic accuracy.
• Statistical analysis suggestions are provided based on the density-amplitude representation.