Smartphone Networked Optical Workflow
SNOW
A lateral-flow test develops colored lines or spots on a membrane. Read by eye, a faint one is a judgement call that two people can make differently and neither can revisit. SNOW makes it a measurement instead.
Three reads, one device
The same cassette, three outcomes. The middle one is the case that matters: a spot faint enough that two people would call it differently.
The control developed and nothing else did. A valid test, and a clean answer.
A single spot barely above the membrane. Called by eye, this is the reading two people disagree about.
Spots developed but the control did not. Reported as invalid rather than as a result, because a control that failed says nothing about the analytes.
Lines, squares, or spots
Most lateral-flow tests develop bands across a strip; some, like the canine tick-borne panels, develop spots on a keyhole membrane. SNOW reads both — it locates each zone, wherever the device puts it, and measures it against the membrane immediately around it. That is what lets one reader serve a whole test menu instead of one cassette.
Reagent-pad strips are next. A urine dipstick carries a row of squares that each change color rather than intensity, so it is read against a reference chart — a different problem from a band, and one a phone camera is well suited to once the color is corrected properly. That work is in development and not yet reading in clinics.
For bands and spots the geometry is all that differs: the reader finds the zone, measures it against the membrane on either side, and returns a number rather than an impression. A dipstick asks a different question — not how dark a pad has gone, but which color it has landed on — and answering it reliably is the reason that format is a program rather than a feature.
In development
Four layers, decoupled on purpose
Each can fail without taking the others down. That is why the reader can be useful in a clinic long before the surveillance layer has enough coverage to say anything.
Capture
Guided photography on an ordinary phone. The cassette type is detected, and a capture that cannot be read — glare across a zone, blur, a window running off the frame — is rejected at the bench while the test is still in hand.
Read
The window is located and straightened, so a cassette photographed at an angle or upside down reads the same as one squared up under good light. Each zone — a band across a strip or a spot on a membrane, the geometry differs and the measurement does not — is then read as optical density against the membrane immediately around it, which is what makes the same device read the same under clinic fluorescents and daylight.
Record
Every read keeps its photograph, what the software measured, and what the clinician concluded. Where those two disagree, both are kept — so there is something to go back to, which a test read by eye does not leave.
Surveil
Readings aggregate into positivity by place and time. Because adoption is clinic-driven, observed counts confound incidence with how much testing is being captured — so coverage is modeled explicitly, and the map says how confident it is rather than presenting sparse data as a signal.
Why a phone can do this at all
A phone is an uncontrolled optical instrument. Sensors differ between models, the image pipeline is proprietary and changes without notice, and clinic lighting is whatever it is. Any approach that depends on absolute brightness fails the first time someone stands near a window.
So nothing here depends on absolute brightness. Each zone is measured against the membrane immediately beside it. Under bright light both the band and the membrane are lighter; under dim light both are darker; the difference between them is what carries the signal, and it survives.
Saying when it cannot tell
A control zone that did not develop invalidates the test, and the software reports that rather than returning a negative. Glare across a zone, a window running off the frame, a photograph too small to hold a band — each is named, with what to do about it.
Disagreement is kept, not overwritten
When a clinician disagrees with a reading, the original measurement is kept alongside their correction. Those disagreements are the most useful records the system holds: they are how the reader is checked, and how it improves.
A rapid test read by eye leaves nothing to review when a case goes sideways. This leaves the photograph, the measurement, and the call.
Current status
SNOW is in development and is not a cleared or approved diagnostic device. The canine tick-borne and heartworm panel reads against an internal reference set of labeled photographs; the feline panel is not yet validated. Every result is presented for a clinician to confirm, and none of it substitutes for their judgement.