Page 76 - Annual report 2021-22

Annual Report 2021-22 |

Viren Sardana

59

Viren Sardana uses electronic sensors, clinical informatics and AI based approaches to study human
physiology and apply the knowledge for biomedical engineering.

In collaboration with Debasis Dash, Viren Sardana was involved in RAPID-CT that developed an AI
system for detection of clinical triage in emergencies.

In India, there is one radiologist per 1 lakh patients. This causes delay in turnaround time in diagnosis
reports for the patients. There is a need for AI based algorithms to augment and expedite the workflow
by automating specific parts of the workflow. The idea behind RAPID-CT was to understand the means
and guidelines necessary for development of AI models in the country. RAPID-CT attempted to provide
methods for cases where an indigenous solution is needed.

Due to the occurrence of the COVID-19 pandemic, the project resources were diverted to meet the
demand for chest X-Ray (CXR) based diagnostic software to identify COVID pneumonia. An AI model
was built which consisted of an ensemble of three models (lung segmentation model, lung opacity
detection model, and pathology detection model) which combine into a rule-based decision-making
algorithm designed by the collaborator radiologists. The accuracy on select data (Indian Quarantine
Center based in Mumbai) was 87% while the negative predictive value was 98%. Data was collected,
acquired, and annotated for COVID-19 detection and localization of COVID-19 on the scans.
He was also involved in Rakshak radiology project for annotation of chest X rays. 2200 CxRs were
annotated from BIMCV and made available for public access. CxRs were also collected from an Indian
quarantine center along with metadata (Pathology reports, Corresponding RT-PCR status, and
Medication Information). Algorithm for COVID-Pneumonia detection using CxRs was developed and
made available for inferencing at the CovBase4all software platform.
In collaboration with Shantanu Sengupta and Debasis Dash; Viren Sardana was instrumental in SARS
CoV2 serology and vaccine response in Phenome India project that collected samples from a
prospective cohort of CSIR-IGIB employees and their family.

Viren Sardana was also involved in a multi-institutional collaboration to create a public health data
informatics portal which would be expandable and adaptable. This portal would organize and share
digital data for public health epidemiology in India to gain an integrative view of the population health
challenges. The overall objective of the program is to carry out analytics over the data, examine and
evaluate data about epidemiology, diseases burden, clinical trials being generated all over the country.
The project has three major components - (a) making health data inventory and repository; (b)
analytics and interpretation; and (c) visualization and generating advisories. Data acquisition would
be enabled by coordinated efforts of the participating institutions/organizations through developing
tools to harmonize and collate the relevant epidemiological health data currently available, with
parallel efforts driven towards streamlining the new data acquisition. For the subsistence of data
repositories development of infrastructure would be required that aids in the maintenance of
centralized secured data housing with established access control policies, which forms the second
objective of the current project. Finally, an advanced data analytics technology platform would be