Prevalence and Clinical Association of Paroxysmal Slowing in Epilepsy
Abstract number :
3.153
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2023
Submission ID :
6
Source :
www.aesnet.org
Presentation date :
12/4/2023 12:00:00 AM
Published date :
Authors :
Presenting Author: Yonatan Serlin, MD, FRCPC – Dalhousie University
Hamza Imtiaz, MSc – Dalhousie University; Tamir Avigdor, MSc – McGill University; Anna Minarik, BSc – Dalhousie University; Sina Lash, BSc – Dalhousie University; Ben Whatley, MD – Dalhousie University; Kristin Ikeda, MD – Dalhousie University; Dan Milikovsky, MD PhD – Tel-Aviv University; Alon Friedman, MD PhD – Dalhousie University
Rationale: Paroxysmal slow wave events (PSWEs) are EEG segments in which the median power frequency falls below 6 Hz for at least five consecutive seconds. A recent study has linked PSWEs to the prediction of epilepsy in patients with a first seizure. We investigated the prevalence and localization of PSWEs in a large cohort of epilepsy patients and individuals without epilepsy to determine if PSWEs can serve as a biomarker for drug-resistant epilepsy (DRE).
Methods: We used natural language processing (NLP) and in-house computer algorithms to extract clinical and EEG data from a subset of 4686 outpatient records available in the Temple University Hospital (TUH) EEG corpus. In addition, we manually examined clinician reports from a random sample of 500 subjects to classify and compare patients with a validated epilepsy diagnosis and subjects without the diagnosis of epilepsy presenting with seizure-mimics who underwent routine EEG interpreted as normal. We further performed a longitudinal study on 30 and 22 pairs of EEGs from patients diagnosed as "seizure-free" or DRE, respectively. All records had at least 17 common channels (10-20 system) with a recording duration of at least 10 minutes.
Results: The occurrence and proportion of time spent in PSWEs was significantly higher in the manually validated epilepsy (n=192) and NLP-detected epilepsy (n=866) groups compared with patients without known epilepsy (n=178), regardless of the number of antiseizure medications used. Topographic heatmaps showed PSWEs to be most abundant in bilateral anterior-temporal head regions. The proportion of patients with prolonged time in PSWEs was significantly higher among DRE patients (p=0.014) and time in PSWEs significantly increased when patients became drug resistant (p=0.013). Time in PSWEs demonstrated moderate to good discriminatory ability for classifying subjects as DRE (AUC=0.704).
Conclusions: This is the first report on quantitative analysis of PSWEs in a large open-source EEG database. Prolonged time in PSWEs in a bilateral anterior-temporal distribution can indicate the lack of response to therapy in patients with epilepsy.
Funding: This study was supported by the Canadian Institute for Health Research (CIHR)-ERA-NET grant #168164 (A.F.); CIHR project grant #180636 (B.W., A.F.); The Israel Science Foundation grant #096409 (A.F.); The Israel-USA binational Science Foundation grant # 2021133 (A.F.).
Neurophysiology