Medical software module autoSCORE™

clinicalEEGneurological

Overview
autoSCORE™ is an AI-enabled EEG analysis application that leverages deep learning to provide expert-level EEG interpretation. It flags clinically relevant abnormalities and delivers study-level insights across long-term monitoring (LTM), ambulatory and routine EEG studies. By automating routine assessments and highlighting key findings, autoSCORE is designed to reduce manual review burden and help clinical teams focus on patient care.

Key benefits

• Expert-level EEG interpretation driven by deep-learning models
• Automatically flags and subclasses detected abnormalities to guide review prioritization
• Provides study-level indicators and visual previews of abnormality prevalence
• Helps accelerate study reviews and may reduce inter-scorer variability

How autoSCORE assists review

• Automatically classifies detected abnormalities into four subtypes: focal epileptiform, generalized epileptiform, focal non-epileptiform, diffuse non-epileptiform
• Provides visual preview of prevalence of abnormalities before opening a study
• Enables quick navigation to abnormal EEG segments ranked by abnormality type and probability score
• Supports review while studies run (useful for ambulatory recordings)

Supported study types

• Long-term monitoring (LTM)
• Ambulatory EEG
• Routine EEG

Workflow integration (example flow)

• Step 1 — Start study with autoSCORE enabled
• Step 2 — Use autoSCORE indications in the study list during recording or upon study completion for workflow planning
• Step 3 — Review the study by sorting autoSCORE findings according to abnormality type and probability
• Step 4 — Achieve faster time to report

Classification and visual indicators
Study-level indicator header: Last | First | Abnormality | Focal Epi | Gen Epi | Focal Non-Epi | Diff Non-Epi
Example row: Edwards | Nancy | 99% | 89% | | 94% | 50%

Study-segments header: Last | First | Abnormality | Focal Epi | Gen Epi | Focal Non-Epi | Diff Non-Epi
Example row: Virotsko | Cynthia | 3/9 | 3/9 | 1/9 | 0/9 | 0/9

Proven performance
Developed using a deep-learning model trained on a dataset of over 30,000 expertly labeled EEG recordings and validated in peer-reviewed publications. Reported performance metrics in validation studies show results on par with leading human experts.

Peer-reviewed validation
Performance results and validation are published in peer-reviewed journals (see cited literature).

System & availability
autoSCORE is developed by Holberg EEG and is available exclusively with Natus NeuroWorks EEG software (version 10 or higher). Not available in all countries.

Specifications

• Product name / model: autoSCORE™
• Integration: Exclusive integration with Natus NeuroWorks EEG software (v10+)
• AI model: Deep-learning model trained on >30,000 labeled EEG recordings
• Supported study types: Routine EEG, Ambulatory EEG, Long-term monitoring (LTM)
• Abnormality subclasses: focal epileptiform, generalized epileptiform, focal non-epileptiform, diffuse non-epileptiform
• Performance (reported validation results): specificity ≈ 90%, sensitivity ≈ 87%, overall accuracy ≈ 88%, normal vs abnormal discrimination ≈ 95%
• Key outputs: study-level abnormality probability, segment-level abnormality probabilities, visual prevalence previews, sortable findings by abnormality type and probability
• Regulatory / availability note: Product availability varies by country