4 ms·
Would it be possible to make a paper one have words (such as pregnant/not pregnant) instead of one/two lines?
by blendergeek 6y ago
Would it be possible to make a paper one have words (such as pregnant/not pregnant) instead of one/two lines?
- philips 6y agoHave an app that uses the phone camera?
- blendergeek 6y agoNo. I'm talking about designing a paper pregnancy test that has words instead of lines.
- maccard 6y agoThat's not an awful idea, and a step forward from the idea of replaceable strips on a reader that the thread suggests!
- pedrow 6y agoI suspect it is difficult to get the 'raw' output from a cell-phone camera so you are dependent on the phone's processing (contrast enhancement etc), which might vary between models. Not impossible (e.g. having control lines) but needs thinking about.
- refurb 6y agoNot for this test. It's a lateral flow assay [1]. The urine is absorbed by the strip and then moves along with capillary action. If there is hCG (human chorionic gonadotropin) in the urine, then the it binds to linked-antibodies (often linked to gold atoms) and drags them along, then those antibodies bind to the actual assay and a control assay (antibodies to the antibodies). I supposed you could make the line words, but they'd be pretty small. [1]https://en.wikipedia.org/wiki/Lateral_flow_test https://en.wikipedia.org/wiki/Lateral_flow_test
- Johnny555 6y agoI don't think that would solve the problem. "The word "not" is faintly showing, what does that mean?" It it was trivial to show a single clear indicator word, they could just show a clear indicator line, then there'd be no ambiguity. It's not the second line that confuses people, it's figuring out if that line is really visible enough.
- srtjstjsj 6y agoBut when the line is faint you get false negative from the machine.
- Johnny555 6y agoDo you? The entire point of the machine is to read the test to the limits of what the test strip can detect. It the digital reader can't detect the line, that should be a limitation of the test strip, not the automation. And it should do better than an untrained human.