detectQS
Neonatal EEG / aEEG clinical review tool
Review now gets a head start: AI-suggested Sleep‑Wake Cyclicity intervals and an automatic PMA estimate, tuned to stay quiet unless confident and never written to a case until you personally confirm it.
The AI's job is to hand you a head start, not a verdict. Every suggestion goes through the exact same confirm step as an interval you'd draw by hand.
The classifier flags candidate Quiet Sleep segments per hemisphere and draws them as a dashed purple region -- deliberately tuned conservative, so it stays quiet rather than flooding the trace with guesses.
Click into a suggestion like any other interval: drag its edges, or reject it outright. Nothing about the case has changed on disk yet.
Hit Confirm and it's saved exactly like a hand-drawn interval -- plus a record of the model's original bounds and confidence, for comparison later.
A live case with two AI-suggested intervals (purple) awaiting review, and a PMA estimate run alongside your Burdjalov subscores.
Retuned after early testing showed it was over-suggesting -- the threshold and minimum run-length were both raised specifically so it doesn't spam you with marginal proposals.
Accepted, adjusted, and now rejected suggestions are all logged with the model's original bounds and confidence -- the full record future model versions learn from.
Your Burdjalov subscores still suggest a PMA range the way they always have -- the AI estimate runs alongside as its own independent read, never in place of your judgment.
Annotations now live in their own annotations/ subfolder inside each case -- copy or zip a case and every reviewer's work travels with it.
Check off any cases in the new Export column and bundle them -- raw signals plus each case's full annotations/ folder -- into one zip.
