Automated epileptiform spike detection via affinity propagation-based template matching

Area of Science:

• Neurology
• Biomedical Signal Processing
• Machine Learning

Background:

• Interictal epileptiform spikes are key diagnostic biomarkers for epilepsy.
• Current visual inspection for spike detection is time-consuming and requires expert neurologists.
• Automated systems are needed for faster and more reliable epilepsy diagnosis.

Purpose of the Study:

• To develop an efficient automated spike detection system for epilepsy diagnosis.
• To improve the accuracy and speed of detecting interictal epileptiform spikes.
• To evaluate template matching techniques for spike detection.

Main Methods:

• Developed an efficient template matching spike detector using combined spike and background waveform templates.
• Generated a template library by clustering spike and background waveforms from 50 epilepsy patients.
• Benchmarked five clustering techniques using receiver operating characteristic (ROC) curves.

Main Results:

• The affinity propagation-based template matching system achieved the highest Area Under the Curve (AUC) of 0.953.
• Integrating background templates with spike templates significantly improved detection performance.
• The proposed method outperformed four other conventional template matching techniques.

Conclusions:

• An efficient template matching system combining spike and background templates can automate epilepsy spike detection.
• This automated approach offers a faster and more reliable alternative to manual visual inspection.
• The affinity propagation clustering method shows strong potential for clinical application in epilepsy diagnosis.