A comparison between detectors of high frequency oscillations

Area of Science:

• Neuroscience
• Biomedical Engineering

Background:

• High-frequency oscillations (HFOs) are critical biomarkers for detecting epileptogenic tissue.
• Manual identification of HFOs is labor-intensive and subjective, necessitating automated solutions.

Purpose of the Study:

• To compare the performance of four automated HFO detection algorithms on a unified dataset.
• To introduce and evaluate a novel HFO detector with an improved feature for handling channels lacking a discernible baseline.

Main Methods:

• Intracerebral EEGs from 20 patients were analyzed, with HFOs and baselines identified by expert reviewers.
• Four existing HFO detectors and a newly developed detector were implemented and configured for optimal performance.
• Performance was assessed using receiver operating characteristic curves, false discovery rates, and channel-level rankings.

Main Results:

• All evaluated HFO detectors showed improved performance when utilizing an optimized configuration.
• The novel detector exhibited higher sensitivity and lower false positive rates compared to other methods, with comparable false detection rates.
• Performance differences were most pronounced in channels with high HFO activity.

Conclusions:

• The developed detector outperformed existing methods on the tested dataset, particularly in challenging channels.
• Optimizing detector configurations for specific data types significantly enhances performance.
• Standardized dataset comparisons are vital for evaluating HFO detection algorithms and understanding validation challenges.