Based on our CRADLE project “Digital Photography of Eyes” and the many individual photos that were taken, the conventional leukocoria (white pupillary reflex) examinations helped us to be successful. In the first place, a patient photographs his eyes more often than a physician, who usually only performs a red reflex test. We started our CRADLE project in 2014 with a university clinic in the US.
A total of 20 medical check-ups were performed on newborns up to the age of two. All photos taken by parents, typically 100 - 1,000 photos per child, could be evaluated to examine leukocoria. Due to the fact that these images varied through several different optical axes and distances, the likelihood that the light, regardless of the location, was appropriately reflected by ocular lesions was increased. However, photographic leukocoria are difficult to identify for laypeople, especially if they are shown with low intensity, low resolution, or only in a small spot in the photo. Parents can now monitor their children through artificial neural networks that operate autonomously on smartphones, since the eyes are now routinely scanned for leukocoria during the child's overall development.
We have incorporated our experience from the CRADLE project into our new SEMIC EyeScan app and developed it further to investigate respiratory diseases in people of all ages. According to the AAO (American Association of Ophthalmology), conjunctivitis (pink eye) appears to come from "aerosol contact with the conjunctiva" or from the mucous membrane of the eye, which is transmitted by droplet infections. Before April 3, 2020, we examined data from 450 corona-infected patients with our deep learning system to determine if they were infected by Covid-19. In this context, we checked with our mobile SEMIC EyeScan app by scanning the eyes in video mode whether our solution could reliably detect the Covid-19 infected conjunctiva.
On April 8th, 2020, with additional use of the YUV color model technology and integration of the gamma compression method, we made further progress after evaluating a total of 70,000 patient data. We can now successfully isolate the following types of diseases with a hit rate of over 95%:
SEMIC EyeScan also enables the detection of:
Sequence of the quick tests:
We use a smartphone to scan a patient's both eyes for about 10 seconds. Our company's proprietary AI will then show you the blood pressure, pulse and body temperature in addition to the type of disease within just 3-5 minutes.
Required smartphone camera sensitivity = 2 Mpx min.Back