01
Blind
The clinician observes, examines, and commits a score before seeing model output.
Human signal firstClinical intelligence / 01 Concept prototype
Make every stroke assessment auditable.
WillisVision is a clinician-first NIHSS workflow that keeps human judgment blind before assistive AI enters the room.
Run the clinical loop
SwiftUI architecture
Local-first boundaries
15 NIHSS items modeled
79 logic tests passing
0 decisions delegated
01 Clinician scores first
02 Assist reveals second
WillisVision NIH Stroke Scale
local01 Blind assessment
Facial Palsy
Ask the patient to show teeth or raise eyebrows and close eyes.
AI remains hidden until your judgment is locked.
INTERACTIVE / TAP THE SCORE SIMULATED DATA ONLY
The method / 02
Blind. Reveal. Reconcile.
A compact interaction model designed to protect independent clinical reasoning while making assistive evidence visible and reviewable.
02
Reveal
Assistive observations appear as evidence—not an answer key or substituted decision.
AI as second reader03
Reconcile
The clinician confirms or revises, preserving the reasoning trail as distinct events.
Auditable by designUnder the hood / 03
Built around the failure modes.
The prototype treats capture, inference, scoring, persistence, and export as explicit service boundaries—because clinical AI becomes trustworthy at the seams.
SwiftUIVisionAVFoundationLocal JSON
WV / SYSTEM MAP prototype boundary
01 ExamStore single source of truth
02ACapturecamera / audio
02BAssessmentmodel protocol
02CExportJSON / HTML
Honest boundary Capture and inference are simulated in this concept build. The scoring model,
state machine, Untestable handling, reconciliation, persistence, and export
paths are implemented.
Take the prototype with you One link.
One link.
The full clinical loop.
Open the live interactive demo on any device. No login. No patient data. No install.
chadwycke.io/willisvisiondemo
SCAN / OPEN / INTERACT