Artificial Intelligence-based Video Analysis to Detect Infantile Spasms

A Machine Learning Approach to Infantile Spasms Recognition in Video Recordings

Infantile spasms are a type of seizure linked to developmental issues. Unfortunately, they are often misdiagnosed, causing delays in treatment. The purpose of this study is to develop a computer program that can reliably differentiate infantile spasms from similar, yet benign movements in videos. This computer program will learn from videos taken by parents of study participants. Quickly recognizing and treating infantile spasms is crucial for ensuring the best developmental outcomes.

Study Overview

• Status: Completed
• Conditions: Infantile Spasms
• Intervention / Treatment: Device: Spasm Vision

• Study Type: Observational
• Enrollment (Actual): 61

Contacts and Locations

Study Locations

• United States
  • Maryland
    • Baltimore, Maryland, United States, 21287
      • Johns Hopkins Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Participants will be recruited from the outpatient and inpatient settings of Johns Hopkins Hospital, an academic medical center located in Baltimore, Maryland offering tertiary and quaternary care.

Description

Inclusion Criteria:

• Participant age less than 24 months
• Participant evaluated in the Johns Hopkins Outpatient Center, Johns Hopkins Pediatric Emergency Department or Johns Hopkins Inpatient Units due to spells of abnormal movement or seizure
• Participant evaluated by a pediatric neurologist during the outpatient or inpatient visit at Johns Hopkins Hospital
• At least one video recording of the spell of abnormal movement produced by the parent/guardian available for provider review

Exclusion Criteria:

• Poor video recording quality
• Entire patient is not in frame

Study Plan

How is the study designed?

Number of groups / cohorts

3

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Confirmed Epileptic Spasms (Positive Class); Participants diagnosed with infantile spasms based upon historical data and supportive electroencephalography data (i.e. hypsarrhythmia or modified hypsarrhythmia background).	Device: Spasm Vision; Machine learning software developed to analyze videos and accurately distinguish infantile spasms from visually similar movements.
Epileptic Spasm Mimics (Negative Class); Participants diagnosed with non-epileptic movements (e.g. Sandifer syndrome, shuddering attacks, stretching, stereotypy, startle reflex, writhing movements, jitteriness, sleep myoclonus) based upon historical data and supportive electroencephalography data (when available).	Device: Spasm Vision; Machine learning software developed to analyze videos and accurately distinguish infantile spasms from visually similar movements.
Awake and Alert (Negative Class); Participants exhibiting spontaneous, subtle movements in the awake and alert state.	Device: Spasm Vision; Machine learning software developed to analyze videos and accurately distinguish infantile spasms from visually similar movements.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Model Sensitivity (Recall)	Proportion of true positives which the model classified correctly in the test dataset.	2 years
Model Specificity	Proportion of true negatives which the model classified correctly in the test dataset.	2 years
Model Positive Predictive Value (Precision)	Proportion of positive classifications which were correct in the test dataset.	2 years
Model Negative Predictive Value	Proportion of negative classifications which were correct in the test dataset.	2 years

Collaborators and Investigators

• Sponsor: Johns Hopkins University
• Investigators: Principal Investigator: Eric Kossoff, MD, Johns Hopkins Neurology; Principal Investigator: Rama Chellappa, PhD, Johns Hopkins Biomedical Engineering

Study record dates

Study Major Dates

• Study Start (Actual): August 26, 2024
• Primary Completion (Actual): May 6, 2026
• Study Completion (Actual): May 6, 2026

Study Registration Dates

• First Submitted: March 11, 2024
• First Submitted That Met QC Criteria: March 11, 2024
• First Posted (Actual): March 18, 2024

Study Record Updates

• Last Update Posted (Actual): May 7, 2026
• Last Update Submitted That Met QC Criteria: May 6, 2026
• Last Verified: May 1, 2026

More Information

Terms related to this study

• Keywords: Epilepsy; Artificial Intelligence; Hypsarrhythmia; West Syndrome; Machine Learning; Seizure; Computer Vision; Infantile Spasm; Video Analysis; Epileptic Spasm
• Additional Relevant MeSH Terms: Epileptic Syndromes; Neurologic Manifestations; Brain Diseases; Central Nervous System Diseases; Nervous System Diseases; Epilepsy, Generalized; Pathological Conditions, Signs and Symptoms; Signs and Symptoms; Epilepsy; Seizures; Spasms, Infantile
• Other Study ID Numbers: IRB00429753

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No