Authors :
Presenting Author: Emily Peter, BS – Boston Children’s Hospital
Michele Jackson, BA – Neurology – Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA; Lillian Voke, BS – Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA; Edeline Jean Baptiste, BS – Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA; Sarah Cantley, BS – Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA; Nathan Hall, MSc – Miku Inc, New Jersey, USA; Eric White, MSc – Miku Inc, New Jersey, USA; Tobias Loddenkemper, MD – Boston Children’s Hospital and Harvard Medical School, Boston, MA, USA
Rationale:
Seizure detection devices may permit earlier seizure detection and provide a greater sense of safety and security for patients and families living with epilepsy; however, prolonged use of wearable devices can be uncomfortable for patients with comorbidities such as autism1. Our team’s recent work demonstrates that video-based seizure detection algorithms are possible2. We aimed to assess the feasibility and challenges of implementing a video, audio, and radar baby monitor for contact-free seizure detection in an inpatient setting.
Methods:
We enrolled patients in the epilepsy monitoring unit (EMU) who had a history of motor seizures. We placed a baby monitor mounted on an IV-pole (MIKU Inc., New Jersey, USA), that recorded video, audio, and radar in participants’ rooms. To minimize obstruction of clinical activities and caregiver interference, the IV-poles were placed at the head of patients’ beds on the opposite side of clinical equipment. Researchers utilized an online dashboard to confirm appropriate camera placement. We excluded patients with aggressive behavior, in acute illness or distress, and/or younger than 30 days. To generate higher data quality for machine learning algorithm development, we further refined our inclusion criteria to exclude patients in Posey beds and selected patients who had a history of tonic, clonic, and tonic-clonic seizures.
Results:
We enrolled 41 (46.3% female, median age: 11.8 years) patients admitted to the EMU from August 2022 to May 2023. One patient had no recorded camera data due to camera malfunction. Of the participants, 33 (80.49%) patients had seizures captured by the camera. We reviewed motor seizures and captured 42 tonic seizures, 6 clonic seizures, and 11 tonic-clonic seizures. We overcame enrollment challenges in the following four areas: hardware (4), clinical environment (3), firmware (2), and patient concerns (1) (see Table 1 for a detailed overview of challenges and solutions).
Conclusions:
Our study demonstrates feasibility to implement a contact-free, multi-modal monitoring system within an inpatient setting. Although enrollment efforts were complicated by multiple challenges, feasible solutions were developed for continued monitoring and deployment. Future steps center around machine learning algorithm development for seizure detection.
1. Black MH et al. The use of wearable technology to measure and support abilities, disabilities and functional skills in autistic youth: a scoping review. Scand J Child Adolesc Psychiatr Psychol. 2020 Jul 2;8:48-69.
2. Yang Y et al. Video-Based Detection of Generalized Tonic-Clonic Seizures Using Deep Learning. IEEE J Biomed Health Inform. 2021 Aug 25;8:2997-3008.
Funding:
Shark Tank Mechanism of the Epilepsy Venture Fund/Epilepsy Foundation of America