Climate change, air pollution and AI – battling tuberculosis (TB)

Image credit: Puwadol Jaturawutthichai | shutterstock.com

Air pollution causes nearly 4.2 million deaths every year, the World Health Organization estimates. Apart from impacting cardiovascular, cerebrovascular and respiratory functions, the high concentration of respirable and fine suspended particulate matter (PM2.5 and PM10) also increase the spread of airborne diseases.

Now a computer science professor, along with her students, is implementing AI to predict how airborne diseases can propagate based on climate conditions, air quality, and population density in an area.

Dr. Nupur Giri
(Image courtesy Microsoft)

“If we can employ AI algorithms to predict the spread of disease and understand how environmental changes, including climatic conditions and external factors like pollution impact the ecology and epidemiology of disease, we can initiate precision public health at a granular level,” explains Dr. Nupur Giri, Professor, and Head of Department of Computer Engineering at Vivekanand Education Society’s Institute of Technology (VESIT), Mumbai.

In 2018, the team started their project to find the correlation between these seemingly disparate factors to be able to predict the propagation of infectious diseases in India. The first disease they chose was Tuberculosis (TB).

With the highest number of cases in the world, TB poses a major public health problem in India. The deadly disease affected an estimated 2.74 million people in India in 2017. India also leads the chart, in the occurrence of Multi-drug Resistant TB (MDR-TB) – driving home the urgency to address the issue.

The Indian government has initiated several programs to eliminate TB by 2025, with the National Strategic Plan (NSP) 2017-2025 setting out interventions to contain the incidence, prevalence, and mortality from TB.

One of the biggest challenges of tackling TB is medicine adherence – patients must take their daily drug religiously for six months else they risk developing drug resistance if they do not follow through. This also makes the job difficult for agencies working on containing the disease as they need to ensure patients are taking their medication.

Dr.Giri hopes that their AI-based model can help predict areas of spread of TB and help agencies tackling the disease focus their efforts in those areas.

To study the impact of environmental conditions, such as climatic factors and pollution on the epidemiology of TB, the team collated data from 725 districts over a 17-year time frame from various sources – climate datasets from Skymet Weather, pollution and air quality datasets from Open Government Data Platform India, tuberculosis dataset from the Central Tuberculosis Division and population data from the national census.

With multiple sources of data and varied frequency, the team spent a few months to smoothen, standardize, and normalize the data. Dr.Giri also won a Microsoft AI for Earth grant, and the team’s pace quickened significantly once they started using Microsoft’s Azure platform and the tools provided as a part of the grant.

“The data science virtual machine was a huge benefit, as it comes preloaded with all the tools required, and data crunching became easier. The Machine Learning Studio allowed us to minimize the time required to develop algorithms and write codes, as it has a drag-and-drop authoring environment,” she reveals.

“The initial results are good, but we are currently testing multiple machine learning and neural network models to improve the accuracy of their prediction for every district in India,” she adds.

Once done, they plan to provide a data visualisation dashboard to help those working on eradicating TB make more informed decisions. Predicting the hotspots for TB is just the beginning. The team is hopeful that once it’s cracked the model, it would be able to employ it for other diseases too.

Article provided courtesy Microsoft.

Be the first to comment

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.