Authors :
Presenting Author: Navaneethakrishna Makaram, PhD, MS – Boston Children's Hospital, Harvard Medical School, Boston, MA 02115, USA
Melissa Tsuboyama, MD – Boston Childrens Hospital
Jeffrey Bolton, MD – Boston Childrens Hospital
Joseph Harmon, MD – University of Utah Health
Christos Papadelis, PhD – Cook Children's Health Care System
Scellig Stone, MD, PhD – Boston Childrens Hospital & Harvard Medical School
Phillip Pearl, MD – Boston Children's Hospital & Harvard Medical School
Alexander Rotenberg, MD, PhD – Boston Children's Hospital, Harvard Medical School, Boston, MA, USA
Ellen Grant, MD – Boston Childrens Hospital
Eleonora Tamilia, PhD – Boston Children’s Hospital, Harvard Medical School, Boston, MA, USA
Rationale:
A significant number of children with drug-resistant epilepsy (DRE) continue to have seizures after resective surgery, indicating a need for enhanced estimators of the epileptogenic zone (EZ). We hypothesized that visually imperceptible, low-entropy activity during interictal periods, even in absence of typical epileptic spikes, is a strong localizing EZ signal. To test this, we mapped "low-entropy-zones" using intracranial EEG (iEEG) in children with DRE. We then assessed how stable these zones remained over extended periods, and whether targeting these zones during surgery predicted favorable postsurgical outcomes.
Methods:
We analyzed iEEG data from 75 children with DRE (Fig 1A). This included short (5-minute) data from 59 patients with known Engel outcomes (used for outcome prediction) and prolonged (3-hour) data from a separate cohort of 16 patients (used to assess stability). We calculated each contact's entropy across various frequencies (broadband, delta, spike, gamma, ripple and fast-ripple, Fig 1B) and identified low-entropy-zones using patient-specific thresholds (Fig 1C, D). We then used 3-fold cross-validation to determine if high low-entropy-zone removal predicted patient outcome (Fig 1E). Predictive value of entropy measures during non-epileptiform (spike-free) epochs was also assessed. Additionally, we tested established interictal estimators (i.e., spikes-on-ripple and fast-ripples) for their ability to predict outcomes. We used F1 measure to quantify the performance. Finally, we investigated whether the entropy distribution observed in brief epochs was consistent with that estimated from prolonged data.
Results:
The low-entropy-zone overlap with resection was correlated with favorable Engel class (p< 0.0001, R=-0.54). This correlation persisted even during non-epileptiform epochs (R=-0.52). The removal of low-entropy-zones predicted outcomes with an F1-score of 87% (p< 0.0001, Engel 1 vs. 3-4; Fig 2A). When only non-epileptiform epochs were used, low-entropy-zones maintained a high predictive value (F1-score=89%, Fig 2B). Low-entropy-zone mapping outperformed spikes-on-ripple (F1-score=82%, p=0.002, Fig 2C) and fast-ripples (F1-score=80%, p=0.01, Fig 2C) in predicting outcome. Interictal entropy distribution over brief epochs showed very strong correlation with prolonged data (R >0.8, p< 0.0001), and its relationship with the seizure onset zone did not significantly differ between brief and prolonged data (p >0.6).
Conclusions:
We present low-entropy-zone as a new accurate and stable interictal EZ estimator in pediatric DRE, which can be measured via brief iEEG data even in absence of epileptiform discharges. The ability to detect low-entropy activity, outside of overt seizures or spikes, allows us to pinpoint epileptogenic regions that might otherwise go unnoticed. Low entropy zones appear to have value in predicting postoperative outcomes in children with drug-resistant epilepsy and outperforms other conventional iEEG markers, such as spikes-on-ripples or fast-ripples.
Funding:
NIH R03NS127044