autoSCORE (V 2.0.0)

Holberg EEG AS developed autoSCORE (V 2.0.0), a Class II software-only artificial intelligence device engineered to aid qualified medical practitioners in reviewing EEG recordings by identifying and categorizing potential abnormalities with associated probability levels. This AI-driven tool acts as a decision support system, validated through comprehensive non-clinical and retrospective clinical studies, demonstrating comparable safety and effectiveness to predicate devices, notably achieving a higher positive predictive value.

Holberg EEG AS has developed autoSCORE (V 2.0.0), a Class II software-only device designed to assist qualified medical practitioners in reviewing electroencephalogram (EEG) recordings. This artificial intelligence model identifies and indicates sections of EEG that may contain abnormalities, providing a probability level for their presence and categorizing them into four predefined types: Focal epileptiform, generalized epileptiform, focal non-epileptiform, and diffuse non-epileptiform. The device integrates with compatible EEG reviewing software and is intended for use in various clinical settings where EEG data is evaluated, for patients aged three months or older for shorter EEGs, and two years or older for longer EEGs. It functions as a decision support tool, complementing visual assessment by users, and does not provide diagnostic conclusions or treatment recommendations.Comprehensive performance validation, encompassing non-clinical software verification and validation, as well as a retrospective clinical study, was conducted for autoSCORE (V 2.0.0). The clinical evaluation assessed the device's performance against human expert consensus and existing predicate devices. The results demonstrated that autoSCORE (V 2.0.0) achieved a higher positive predictive value (PPV) compared to the primary predicate device and a similar PPV to its previous version. The evaluation concluded that the device's safety and effectiveness profile is comparable to legally marketed predicate devices, affirming that it performs as well and introduces no new safety or effectiveness concerns for its intended use.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. What specific AI algorithms or methodologies does autoSCORE utilize to identify and categorize EEG abnormalities?; How does autoSCORE's performance (e.g., sensitivity, specificity, negative predictive value) compare to human expert consensus and predicate devices across different abnormality types?; What are the limitations of autoSCORE regarding patient age ranges and EEG recording durations, and how might these impact its utility in broader clinical practice?; Could autoSCORE's probability levels and categorization assist in prioritizing complex EEG reviews or aid in the training of new EEG readers?; Are there plans for prospective studies or real-world clinical integration trials to further assess autoSCORE's impact on diagnostic efficiency and patient outcomes?