Research in the Journal of Medical Internet Research
introduced Biocogniv’s new AI-Covid software that can quickly predict the
A number of researchers from the University of Vermont and
Cedars-Sinai found larger accuracy in predicting the probability of COVID-19
infection making use of ordinary blood tests, that help hospitals decrease the
number of patients referred for PCR testing.
Lead author and University of Vermont Assistant Professor
Timothy Plante, M.D., M.H.S stated, "9 months into this pandemic, we today
have a clear comprehension of exactly how to care for patients with COVID-19,
but there is still a huge bottleneck in COVID-19 diagnosis with PCR
PCR testing is the present standard diagnostic for COVID-19
and requires certain sampling, such as a nasal swab, and then laboratory equipment
Tanya Kanigan, Biocogniv Chief Operating Officer, Ph.D.,
said "According to data from more than 100 US clinics, the national
average turnaround period for COVID-19 testing purchased in emergency rooms is
above 24 hours, much from the focused one-hour turnaround,". Complete
Blood Count along with Complete Metabolic Panels are popular lab tests
purchased by emergency departments and have a fast turnaround time. These tests
offer insight directly into the immune system, kidney, electrolytes, and liver.
The scientists had the ability to teach a design that analyzes changes in these
routine tests and assigns a probability of the patient being COVID-19 negative
with high accuracy.
AI-COVID takes seconds to produce its informative result
when these blood tests return, they may then be integrated by the laboratory
into a test interpretation."
The Biocogniv staff thinks a secondary advantage of
laboratories incorporating AI-COVID may be reduced time for conventional PCR
The AI-COVID model was validated on real-world data offered
by Cedars Sinai and on data from geographically and demographically different
patient encounters from 22 U.S. hospitals, achieving a location underneath the
curve of 0.91 of 1.00.
This enables the design to attain a high sensitivity of 95%
while maintaining reasonable specificity of 49%, and that is pretty similar to
the overall performance of other widely used rule-out tests.