Sleep Tool · Clinical Validation
Clinical validation
Automated scoring should not cost a clinic diagnostic confidence. Aegis tests the Sleep Tool against expert scoring, with sleep physicians in the loop.
Method
What we measure
Each measure answers a question a sleep physician would ask.
Epoch-by-epoch agreement
How often the tool's label for a 30-second segment matches the expert's label.
AHI correlation
How closely the tool's apnea-hypopnea index tracks the index from physician-scored studies.
Time to draft report
How long it takes to go from the raw recording to a report ready for review.
The benchmark
Experts are the reference, and experts disagree
Sleep scoring involves judgement. Two experienced scorers rarely match on every segment, so the fair target is the level of agreement between experts, not perfection.
The data science team has research experience in label noise and disagreement between raters. That shapes how the model is trained and how results are reported.
Clinical oversight
Led by a sleep physician
Aegis's Chief Medical Officer is board-certified in internal medicine, intensive care medicine and sleep medicine, and a Fellow of the American Academy of Sleep Medicine. He leads weekly validation sessions with clinical partners and checks that the tool's output meets sleep medicine standards.
The model has been developed with Canadian sleep centres. 4,000 patient files have been ingested for training.
Results
Validation data on request
We share results with clinics, partners and investors together with the study design behind them. Technical documentation is available on request.
Intended use
The Aegis Sleep Tool is not currently offered for independent commercial diagnostic use. It is limited to validation, pilot evaluation and investigational research while Health Canada MDEL and FDA requirements are completed.
Questions
Validation questions
How is the Aegis Sleep Tool validated?+
By comparing its scoring with studies scored by experts, using measures such as epoch-by-epoch agreement, correlation of the apnea-hypopnea index (AHI), and time to a draft report. Validation sessions with clinical partners are held weekly and led by Aegis's Chief Medical Officer.
Why compare against experts rather than a single correct answer?+
Two expert scorers do not always agree on every 30-second segment. Agreement between experts is therefore the realistic benchmark for an automated tool.
Does Aegis publish accuracy figures?+
Results are shared with clinics, partners and investors on request, with the study design that produced them. Aegis does not publish headline figures on this site until the validation study is complete.
How do I request validation data?+
Email technical@aaegis-ai.com.
Request validation data
Tell us who you are and what you need to evaluate. The technical team will reply with the material that applies.