The Effect of AI-assisted cEEG Diagnosis on the Administration of Antiseizure Medication in Neonatal Seizures

The Effect of AI-assisted cEEG Diagnosis on the Administration of Antiseizure Medication in Neonatal Seizures

AI-assisted cEEG Diagnosis of Neonatal Seizures to Optimize the Administration of Antiseizure Medication: a Multicenter, Randomised, Controlled Trial

Chengdu Women's and Children's Central Hospital, Xiamen Children's Hospital, Kunming Children's Hospital, The Affiliated Hospital Of Southwest Medical University, Children's Hospital of Zhengzhou University

This is a prospective randomised clinical trial study to test an artificial intelligence (AI)-assisted continuous electroencephalogram(cEEG) diagnostic tool for optimizing the administration of antiseizure medication (ASM) in neonatal intensive care units(NICUs).

The occurrence of neonatal seizures may be the first, and perhaps the only, clinical sign of a central nervous system disorder in the newborn infant. The promoted treatment of seizures can limit the secondary injury to the brain and positively affect the infant's long-term neurological development. However, the current antiseizure medication (ASM) are both overused and underused. Studies indicated that early automated seizure detection tool had a high diagnostic accuracy of neonatal seizures. However, there is little evidence that early automated seizure detection tool could the optimize the administration of ASM and improved the neurological outcomes in neonatal seizures. Therefore, the primary study aim is to investigate whether the utility of AI assisted cEEG diagnostic tool could optimize the administration of ASM in NICUs. This project will enroll the neonates with suspected or high risk of seizures who will receive at least 72 hours cEEG monitoring during hospitalization. All the cEEG monitoring methodology is standardized across recruiting hospitals. The intervention will be an artificial intelligence (AI)-assisted continues electroencephalogram (cEEG) diagnostic tool. The individuals were randomly allocated to one of the two groups using a predetermined randomisation sequence and block randomisation generator (block of 4). The group 1 will be monitored with cEEG and the cEEG recording will be assessed by neonatologists with AI assisted cEEG diagnostic tool in real time during cEEG monitoring. The group 2 will be monitored with cEEG and the cEEG recording will be assessed by neonatologists when as routine during cEEG monitoring. Both groups will follow the standard clinical protocols for ASM administration of the recruiting hospitals The reference standard is the electrographic seizures interpreted by 3 clinicians who had attended the uniformly training program and were certified by the Chinese Anti-Epilepsy Association. These 3 clinicians are blinded to the group allocation.

This group will be monitored by cEEG with standard operating procedure. The cEEG recording will be evaluated by neonatologists with the routine assessment protocol and AI assisted cEEG diagnostic tool in real time during cEEG monitoring. Both real-time cEEG and amplitude-integrated EEG traces are displayed at the bedside for clinical review. This group will follow the standard clinical protocols of the recruiting hospitals for ASM administration after the neonatologists' review.

This group will be monitored by cEEG with standard operating procedure. The cEEG recording will be evaluated by neonatologists with the routine assessment protocol during cEEG monitoring. Both real-time cEEG and amplitude-integrated EEG traces are displayed at the bedside for clinical review. This group will follow the standard clinical protocols of the recruiting hospitals for ASM administration after the neonatologists' review.

The AI-assisted cEEG diagnostic tool is an automated seizure reporting system, including a quantitively EEG neural signal processing pipeline to extract features from the original signal datasets, machine learning models based on gradient boosted model for prediction. The tool can report electrographic seizures in real time during cEEG monitoring. The neonatologists will evaluate the neonates by AI-assisted cEEG diagnostic tool, clinical conditions, real-time cEEG and amplitude-integrated EEG traces. The investigators will make a decision after review the neonates clinical conditions, AI-assisted cEEG diagnostic report, the cEEG and amplitude-integrated EEG.

The routine assessment protocol is that the neonatologists will evaluate the neonates by clinical conditions, real-time cEEG and amplitude-integrated EEG traces.

The inappropriate administration of ASM is defined: (1) the administration of an ASM before the electrographic seizure episode; or (2) an ASM is given to the neonates without electrographic seizure episode.

The GDS comprise comprehensive checklists for assessing neuromotor wholeness, functional maturity, and mental development of infants and toddlers from the perspectives of adaptability, large exercise, fine motor skills, language, and personal-social networking. The GDS score provides an objective assessment of neurological and mental development in this age group.

Total electrographic seizure times per hour (second/hour) is defined as total duration of all seizures in every hour from the start of the EEG monitoring to the end of the cEEG monitoring.

Inclusion Criteria: Postnatal age < or = 28 days; cEEG monitoring at least 24hours monitoring; Suspected seizures； Abnormal movement; Brain infarction； Risk of Intracranial hemorrhage; Abnormality of brain MRI or ultrasound; Hypoxic-ischemic encephalopathy or suspected Hypoxic-ischemic encephalopathy; Central nervous system (CNS) or systemic infections; Suspected genetic diseases or Positive genetic diagnoses; Exclusion Criteria: The neonates with head scalp defect, scalp hematoma, edema and other contraindications which are not suitable for cEEG monitoring during hospitalization.

Rennie JM, de Vries LS, Blennow M, Foran A, Shah DK, Livingstone V, van Huffelen AC, Mathieson SR, Pavlidis E, Weeke LC, Toet MC, Finder M, Pinnamaneni RM, Murray DM, Ryan AC, Marnane WP, Boylan GB. Characterisation of neonatal seizures and their treatment using continuous EEG monitoring: a multicentre experience. Arch Dis Child Fetal Neonatal Ed. 2019 Sep;104(5):F493-F501. doi: 10.1136/archdischild-2018-315624. Epub 2018 Nov 24.

Shellhaas RA, Chang T, Tsuchida T, Scher MS, Riviello JJ, Abend NS, Nguyen S, Wusthoff CJ, Clancy RR. The American Clinical Neurophysiology Society's Guideline on Continuous Electroencephalography Monitoring in Neonates. J Clin Neurophysiol. 2011 Dec;28(6):611-7. doi: 10.1097/WNP.0b013e31823e96d7. No abstract available.

Hoodbhoy Z, Masroor Jeelani S, Aziz A, Habib MI, Iqbal B, Akmal W, Siddiqui K, Hasan B, Leeflang M, Das JK. Machine Learning for Child and Adolescent Health: A Systematic Review. Pediatrics. 2021 Jan;147(1):e2020011833. doi: 10.1542/peds.2020-011833. Epub 2020 Dec 15.