Newswise – Outbreaks of zoonotic diseases transmitted from animals to humans are increasing around the world due to climate change. In particular, the spread of mosquito-borne diseases is very sensitive to climate change, and Chinese Taiwan has seen a worrying rise in the number of dengue fever cases in recent years.
As with most known diseases, the popular adage “prevention is better than cure” also applies to dengue fever. With the world still lacking a safe and effective vaccine for everyone, efforts to prevent dengue fever rely on limiting the places where mosquitoes can lay their eggs and giving people early warning when an outbreak is likely is. So far, however, there are no mathematical models that can accurately predict the location of dengue fever outbreaks in advance.
To address this problem, a research team including Professor Sumiko Anno from Sophia University, Japan, attempted to combine artificial intelligence (AI) with remote sensing data to predict the spatio-temporal distribution of dengue fever outbreaks in Chinese Taiwan. This work, published in Geo-spatial Information Science, was co-authored by Hirakawa Tsubasa, Satoru Sugita, and Shinya Yasumoto, all from Chubu University, Ming-An Lee from National Taiwan Ocean University, and Yoshinobu Sasaki and Kei Oyoshi from the Japan Aerospace Exploration Agency (JAXA), Japan.
First, the team collected climate data of Chinese Taiwan from 2002 to 2020, including data on precipitation, sea surface temperature and shortwave radiation. They also collected information on the place of residence of all reported dengue fever cases registered with China’s Taiwan Center for Disease Control. This allowed the researchers to create a labeled training dataset for the AI model, which should ideally be able to find hidden patterns between dengue fever cases and climate parameters.
The AI model in question was a Convolutional Neural Network (CNN) with a U-Net-based encoder-decoder architecture. “The U-Net model works with remarkably few training images and provides a more precise semantic segmentation with the location information,” says Prof. Anno, explaining the choice of the AI model for her study. This well-established design usually works well for image segmentation tasks, even when trained with few examples. After training the model, the team attempted to validate it using the remaining collected data.
Unfortunately, the model didn’t work as well as the researchers had hoped. Most of the pixels on the map of Taiwan marked as predicted dengue outbreak locations did not match the original data. However, all hopes for this approach are not yet lost, as Prof. Anno points out: “While most of the predicted eruption pixels did not overlap with reality, some of them were quite close to the actual eruption locations. This implies that spatiotemporal prediction of dengue fever outbreaks is possible using remote sensing data.”
Despite the low accuracy of the AI model, this study has shed light on some of the current challenges in using remote sensing data to predict the spatio-temporal distribution of zoonotic outbreaks. The research team believes that using a different model architecture, finding a way to balance the training dataset, and collecting higher-resolution satellite data could be promising ways to achieve the required performance.
More work will be needed before we can use machine learning as a tool to pinpoint potential disease outbreak zones based on climate data, but we must not waver. “Spatiotemporal visualizations generated by deep learning models could potentially guide the implementation of effective measures against disease outbreaks at the optimal time and place for disease prevention and control,” concludes Prof. Anno optimistically.
Let’s hope that further studies in this area will help us protect humans from zoonotic diseases soon.
