And now an AI-based software tool for automated diagnosis of COVID-19 lung infection

A new software tool that revealsthe severity of lung infections in COVID-19 patients has beendeveloped by researchers from the Indian Institute of Science,in collaboration with Oslo University Hospital and Universityof Adger in Norway.

It has been described in a recent study published in thejournal ''IEEE Transactions on Neural Networks and LearningSystems'', Bengaluru-based IISc said in a statement.

COVID-19 can cause severe damage to the respiratorysystems, especially the lung tissues. Image-based methods suchas X-ray or CT scans can prove helpful in determining how badthe infection is, the statement noted.

The software tool, developed by the Departments ofComputational and Data Science and Instrumentation and AppliedPhysics at IISc, called AnamNet, can 'read' the chest CT scansof COVID-19 patients, and, using a special kind of neuralnetwork, estimate how much damage has been caused in thelungs, by searching for specific abnormal features, it said.

''Such a tool can provide automated assistance to doctorsand therefore help in faster diagnosis and better managementof COVID-19'', according to IISc.

AnamNet employs deep learning and other image processingtechniques, which have now become integral to biomedicalresearch and applications. The software can identify infectedareas in a chest CT scan with a high degree of accuracy, itsaid.

The researchers trained AnamNet to look for abnormalitiesand classify areas of the lung scan as either infected or notinfected - this is called segmentation.

The tool can judge the severity of the disease bycomparing the extent of infected area with healthy area.

''It basically extracts features from the chest CT imagesand projects them onto a non-linear space (a mathematicalrepresentation), and then recreates the (segmented) image fromthis representation. This is called anamorphic imageprocessing,'' explains Naveen Paluru, first author and PhDstudent in the lab of Phaneendra Yalavarthy, AssociateProfessor at CDS.

The study also compared AnamNets performance with otherstate-of-the-art software tools which perform similar tasks.

It not only matched its peers in its accuracy, but alsoperformed just as well using fewer parameters. The neuralnetwork was also computationally less complex, which allowedthe researchers to train it much faster to detect anomalies,it was stated.

Another significant advantage of AnamNet is that thesoftware is lightweight with a small memory footprint. Thishas enabled the team to develop an app called CovSeg that canbe run on a mobile phone and hence potentially be used byhealthcare professionals.

''We felt the need for a lightweight framework that couldbe deployed as a point-of-care diagnostic device onsmartphones or a Raspberry Pi,'' says Paluru.

He adds that this feature is missing from currentlyavailable state-of-the-art technologies such as UNet, whichrequires specialised hardware.

According to the authors, AnamNet holds promise beyondmerely identifying lung infections in COVID-19 patients.

''We are currently focusing on making our software morerobust to handle COVID-19 scans, but we are also looking todiversify to other common lung diseases like pneumonia,fibrosis and even lung cancer in the near future, Yalavarthysays.

He suggests that with some changes to the present design,the software could even be used to read brain scans.

The software tool is freely available to the public, IIScadded.