Pigmentation
Cameras can behave differently across skin tones and lighting [2]. That is a real risk. I have not finished a fairness evaluation with labeled subgroups, so I do not claim demographic fairness.
Research
Responsible design
When software touches health, I want uncertainty visible. A system that sounds sure and is wrong is worse than one that admits what it does not know.
Lighting and cameras change what is visible. Photograph for context, not a clinic case.
What I refuse to claim
Lighting and cameras
Glare, exposure, blur, and framing change what landmarks are visible. Device differences still matter after a quality pass. I have not proven the system works across cameras.
Abstention
If confidence is below my current research threshold, I return Unable to assess rather than force a shape label. That behavior exists in code. It does not prove the cutoff is clinically right.
No diagnosis
I measure finger shape linked to digital clubbing, for research review. I do not diagnose primary ciliary dyskinesia or any other disease. Results stay at a shape state, or Unable to assess.
This website stores no patient data. I have not established HIPAA compliance. Bench tests are not human-subjects research.