A machine-learning algorithm for neonatal seizure recognition: a multicentre, randomised, controlled trial

Abstract

Background Despite the availability of continuous conventional electroencephalography (cEEG), accurate diagnosis of neonatal seizures is challenging in clinical practice. Algorithms for decision support in the recognition of neonatal seizures could improve detection. We aimed to assess the diagnostic accuracy of an automated seizure detection algorithm called Algorithm for Neonatal Seizure Recognition (ANSeR). Methods This multicentre, randomised, two-arm, parallel, controlled trial was done in eight neonatal centres across Ireland, the Netherlands, Sweden, and the UK. Neonates with a corrected gestational age between 36 and 44 weeks with, or at significant risk of, seizures requiring EEG monitoring, received cEEG plus ANSeR linked to the EEG monitor displaying a seizure probability trend in real time (algorithm group) or cEEG monitoring alone (nonalgorithm group). The primary outcome was diagnostic accuracy (sensitivity, specificity, and false detection rate) of health-care professionals to identify neonates with electrographic seizures and seizure hours with and without the support of the ANSeR algorithm. Neonates with data on the outcome of interest were included in the analysis. This study is registered with ClinicalTrials.gov, NCT02431780. Findings Between Feb 13, 2015, and Feb 7, 2017, 132 neonates were randomly assigned to the algorithm group and 132 to the non-algorithm group. Six neonates were excluded (four from the algorithm group and two from the non-algorithm group). Electrographic seizures were present in 32 (25•0%) of 128 neonates in the algorithm group and 38 (29•2%) of 130 neonates in the non-algorithm group. For recognition of neonates with electrographic seizures, sensitivity was 81•3% (95% CI 66•7-93•3) in the algorithm group and 89•5% (78•4-97•5) in the non-algorithm group; specificity was 84•4% (95% CI 76•9-91•0) in the algorithm group and 89•1% (82•5-94•7) in the non-algorithm group; and the false detection rate was 36•6% (95% CI 22•7-52•1) in the algorithm group and 22•7% (11•6-35•9) in the non-algorithm group. We identified 659 h in which seizures occurred (seizure hours): 268 h in the algorithm versus 391 h in the nonalgorithm group. The percentage of seizure hours correctly identified was higher in the algorithm group than in the non-algorithm group (177 [66•0%; 95% CI 53•8-77•3] of 268 h vs 177 [45•3%; 34•5-58•3] of 391 h; difference 20•8% [3•6-37•1]). No significant differences were seen in the percentage of neonates with seizures given at least one inappropriate antiseizure medication (37•5% [95% CI 25•0 to 56•3] vs 31•6% [21•1 to 47•4]; difference 5•9% [-14•0 to 26•3]). Interpretation ANSeR, a machine-learning algorithm, is safe and able to accurately detect neonatal seizures. Although the algorithm did not enhance identification of individual neonates with seizures beyond conventional EEG, recognition of seizure hours was improved with use of ANSeR. The benefit might be greater in less experienced centres, but further study is required.

Multichannel EEG abnormalities during the first 6 hours in infants with mild hypoxic–ischaemic encephalopathy

“…of HIE in the primary study 65 40 59 62 Total mild HIE in the primary study 29 24 36 30 No. in the current study (mild HIE and EEG recording <6 hours of life) 9 16 19 14 Definition of mild HIE ≥1 abnormal finding on modified Sarnat Score at 24 hours of life but not meeting criteria for moderate or severe HIE Worst grade of encephalopathy in the first 24 hours using modified Sarnat Score but not meeting criteria for TH Worst grade of encephalopathy in the first 24 hours using modified Sarnat Score but not meeting criteria for TH 1 or more abnormal neurological finding on modified Sarnat Score at 1 hour of life but not meeting criteria for TH Exclusion from pooling (1) Infants who received TH ( n = 1) (1) Infants who received TH ( n = 4) (2) Other diagnosis than mild HIE (1) Infants who received TH ( n = 2) Main publications from cohorts to date Murray et al 57 Murray et al 58 Murray et al 59 Murray et al 30 Murray et al 23 Murray et al 15 Walsh et al 60 Walsh et al 61 Walsh et al 62 Walsh et al 63 Looney et al 64 O’Sullivan et al 65 Rennie et al 66 O’Sullivan et al 67 Finder et al 16 Pavel et al 68 TH therapeutic hypothermia, HIE hypoxic–ischaemic encephalopathy, GA gestational age, CPR cardiopulmonary resuscitation, PA perinatal asphyxia. a Exclusion from primary studies: infants <36 + 0 weeks’ gestation; known genetic disorders and/or inborn errors of metabolism; confirmed sepsis (positive blood or CSF cultures).…”

Neonatal seizures reach the mainstream: The ILAE classification of seizures in the neonate

“…The challenge of eliminating the duration from the neonatal seizure definition is that it hinders our ability to directly apply the current literature to bedside clinical care; recent studies have consistently defined an EEG seizure as at least 10 seconds in duration. [6][7][8][9][10][11] We do not know how the published data apply to seizures that last less than 10 seconds. This change will also introduce potentially significant variability to future studies.…”

Precision Medicine in Neonates: A Tailored Approach to Neonatal Brain Injury

“…Much effort was made to develop automated seizure detection. When tested, all algorithms showed clinically relevant detection rates ( 38 ). Doyle et al ( 39 ) and Malarvilli and Mesbah ( 40 ) used heart rate variability to detect seizures.…”

“…Recently, an ML algorithm for neonatal seizure recognition, ANSeR, was investigated in a randomised controlled trial and was found to be safe and able to detect neonatal seizures. However, it did not yet improve the identification of individual neonates with seizures ( 38 ).…”