Local startup DetectNow is using AI to detect Covid-19. The startup, registered with National Incubation Center(NIC), uses AI to detect the presence of coronavirus by analyzing cough audios.
The startup wanted to offer cost-free screening without delays for coronavirus. Cough is one of the main symptoms of COVID-19 and based on the idea of cough analysis via AI, DetectNow aims to find the unique biosignals of the Covid-19 cough and calculate the probability of a SARS-CoV-2 infection.
The service is available through a web application, making it easy, convenient, and accessible even in low-resource areas, however, the diagnostics function is not live yet, as this will require a medical trial first.
The application has collected 600 samples so far.
DetectNow want to act as an additional tool in fighting Covid-19 by assisting national health care systems in the screening, outbreak prevention, and outbreak control process.
“We are awaiting verified data from COVID-19 positive tested people. This has to be done in a clinical trial to protect Data privacy,” the team behind the initiative said in a statement.
“Considering that we want to CLEAN a drum recording (highly explosive sounds), we removed low frequencies below 40 Hz and high frequencies above 15 kHz – 18 kHz because of microphone limitations. We also used a gate to remove unwanted noise between the coughs and make sure that you do not remove the “silent coughs” or “heavy breathing” because this might probably be important features for the model. Finally normalizing the sound, so that we do not have different loudness/ amplitudes among the files.”
The team hopes to classify cough sounds into two main categories; Covid-19 cough (dry cough) and Non-Covid-19 cough based on different cough patterns.
DetectNow are a non-profit, work for the social good, and are humanitarian and plan to stay that way. They hope their approach to cough analysis will provide a foundation towards further clinical research with AI on pulmonary diseases.