This AI-based software tool can diagnose Covid-19 lung infection

The tool can judge the severity of the disease by comparing the extent of infected area with healthy area

When 44-year-old Arvind B (name changed) was diagnosed with breathlessness in the middle of last year, he found that he could not access ICU care immediately because the radiologist who had to assess his chest scan for severity was delayed.

Unlike some countries which use Artificial Intelligence (AI) to help rapidly assess chest CT scans to determine how seriously a Covid-19 patient’s lung has been compromised, hospitals in India largely rely on traditional human interpretation. This sometimes leads to delays in hospitalisation.

It was a problem that researchers from the Indian Institute of Science (IISc), in collaboration with Oslo University Hospital and the University of Agder in Norway sought to address. Their solution is a new software which can reveal the severity of lung infections of people with Covid-19.

Associate Professor Phaneendra Yalavarthy of IISc’s Department of Computational and Data Sciences (CDS) said the software, which is called AnamNet, uses a special neural network to estimate how much damage has been caused in the lungs, by searching for specific abnormal features and classify areas of the lung scan as either infected or not infected.

CORONAVIRUS SPECIAL COVERAGE ONLY ON DH

Naveen Paluru, first author and PhD student, explained that AnamNet can judge the severity of the disease by comparing the extent of the infected area with the healthy area. “It basically extracts features from the chest CT images and projects them onto a mathematical representation, and then recreates the image from this representation. This is called anamorphic image processing,” Paluru said.

In the case of Covid-19, the software looks for ground-glass opacities which is a spider-web-like abnormality that can form in the lungs of affected patients and pleural effusion where the boundaries of the lung will show as a double boundary or more.

The development of the model began in May 2020. “We can tune it for other respiratory problems, such as lung fibrosis, pneumonia and perhaps even lung cancer by looking for abnormalities associated with those illnesses,” Dr Yalavarthy said.

Dr Anoop Amarnath, a member of the state’s Critical Care Support Unit (CCSU), explained that several modalities, such as clinical, radiological, lab and diagnostics with RT-PCR are considered before a decision is made to move patients to ICU care.

“AI tools can definitely help the radiologists in terms of identifying severe Covid. At the end of the day, it is still the call of the radiologist,” Dr Amarnath said.

Dr Yalavarthy said the software’s small memory footprint has enabled the team to develop a mobile phone app called CovSeg. The software was released into the public domain this week.