Direct Cortical Stimulation and the Irritative Zone in Refractory Focal Epilepsy: A Machine Learning Approach
Abstract number :
1.277
Submission category :
3. Neurophysiology / 3E. Brain Stimulation
Year :
2024
Submission ID :
698
Source :
www.aesnet.org
Presentation date :
12/7/2024 12:00:00 AM
Published date :
Authors :
Presenting Author: Fawzi Babtain, MBBS, MHSc, FRCPC, CSCN (EEG, EMG) – King Faisal Specialist Hospital and Research Centre- Jeddah
Mohammed Al Mansour, MBBS – King Faisal Specialis Hospital and Research Center-Jeddah
Afnan Al Khotani, MBBS – King Faisal Specialis Hospital and Research Center-Jeddah
Rawan Daghistani, MBBS – King Faisal Specialis Hospital and Research Center-Jeddah
Mohammed Bin Mahfoodh, MBBS, FRCS(C) – King Faisal Specialis Hospital and Research Center-Jeddah
Saleh Baeesa, MD – King Faisal Specialist Hospital and Research Centre
Youssof Al Said, MBBS, FRCP(C) – King Faisal Specialis Hospital and Research Center-Jeddah
Rationale:
Understanding the irritative zone (IZ) in SEEG is vital for mapping brain activity and guiding surgery in refractory focal epilepsy (RFE). Its relation to the epileptic network might be challenging, for which we utilized machine learning and direct cortical stimulation (DCS) to illuminate this association.
Methods:
We identified IZ outside and within the epileptic network in spontaneous seizures and DCS, built machine-learning models to classify these zones, and compared the features of each model.
Results:
We stimulated 76 electrodes with 2354 stimulation points. We trained 24 models in the unsupervised machine learning model and identified 3 clusters in each K-means model. Figure 1 shows the principal component analysis (PCA) for both models. The irritative zone outside the epileptic network was most prevalent in Cluster 1 (4%), which demonstrated the highest values for pulse intensity (6.6 mA), actual pulse delivered (5.4 mA), and pulse frequency (44 Hz). Additionally, Cluster 1 showed the highest occurrence of after-discharge of 2-5 Hz at 54%, along with the highest percentage of inducing auras (33%) and habitual signs (44%). On the other hand, the irritative zone within the epileptic network was more prominent in Cluster 0 (15% ), with high-intensity stimulation (6.6 mA mean pulse intensity and 5.4 mA actual pulse delivered), with a mean stimulation frequency of 45 Hz. After-discharges exceeded 10 seconds in 35%. Additionally, 32% of cases report auras, and 44% experience habitual signs following stimulation. No significant differences can be appreciated between the features of each cluster that classified irritative zone irrespective of the epileptic network (figure 2).
Conclusions:
The machine learning approach revealed distinct classification models to identify irritative brain regions with similar clinical and stimulation-related characteristics, regardless of their location. These findings may enhance surgical planning by considering IZ, but further research can augment our findings.
Funding: None
Neurophysiology