Artificial Iintelligence (AI) is being used by South African scientists to detect and classify pollen spores in real time, which may improve forecasting of bioaerosols that trigger respiratory illnesses like allergic rhinitis (hay fever) and asthma.
Prof Jonny Peter, head of the Division of Allergology and Clinical Immunology at Groote Schuur and the UCT Lung Institute, who also leads the South African Pollen Network (SAPNET), says the monitoring of airborne pollen provides an important source of information for the globally increasing number of hay fever and asthma sufferers.
“The worldwide prevalence of allergic rhinitis and asthma rises dramatically each year, both in developed and developing countries. In South Africa, nearly 30% of people suffer with allergic rhinitis and a recent study from Kwazulu-Natal found that 13,7% of adolescents suffer from asthma – of which 9% suffer from severe asthma.
Up to now, scientists have manually counted airborne pollen and spore types using volumetric air samplers that suck air from the environment, collecting the particles it traps on adhesive tape, which are then analysed under a microscope by a pollen expert.
However, the process is cumbersome and costly. With the latest developments in image recognition methods and machine learning, automating this process has become feasible.
Prof Peter says by combining cutting-edge technologies, like AI and imaging flow cytometry, which measures the size, count, shape and structure of a cell, they are able to build a system for South Africa that is capable of identifying and categorising pollen more accurately and at much faster rates.
“In addition to AI providing a more comprehensive picture of pollen in the present, it can also help to model historic environmental change.
“Pollen grains from different plant species are unique and identifiable based on their individual features. Analysing which pollen grains are captured in samples from sediment cores in lakes can help scientists get a better grasp on which plants were thriving at any given point in history, potentially dating back thousands to millions of years. Its numerous applications are transforming the way scientists conduct research and is enabling new discoveries across fields – accelerating scientific productivity.”
Two AI counters have already been set up in Pretoria and Kimberley with two more in Cape Town and Durban coming on line soon. Bloemfontein and Johannesburg will follow in the next few weeks.
“It works similar to the volumetric spore trap that sucks the pollen onto the cellulose membrane, but then it takes a photo of it,” Prof Peter explains. “From there, AI takes over.
“The AI is trained using large volumes of images to recognise the individual grains on the sticky tape. The only challenge is that we need to have a period of validation, so for a time, we will put a conventional volumetric trap next to the automatic AI counter to verify the accuracy of the analysis.
“Once the AI has been sufficiently trained to recognise grains that are indigenous to South Africa, it will be able to classify species very quickly. Main allergens, like grass, plane- and cypress trees, as well as fungal species, like Alternaria, Aspergillus and Cladosporium, should be easy for the AI to identify as these are commonly found around the world,” he adds.
“Pollen counts are often difficult to predict during spring when counts are highest, and the weather is changing daily. Having access to reliable, real-time pollen readings will significantly improve forecasting and, in turn, help hay fever sufferers to mitigate bothersome symptoms that include a runny or itchy nose, sneezing, itchy and watery eyes, as well as nasal congestion.”
2023 pollen forecast with El Niño thrown into the mix
Prof Peter says this year’s pollen season depends largely on the weather, but that hay fever sufferers can expect higher than normal grass counts. “We are already seeing high tree counts in Cape Town with exceptionally high tree counts in Bloemfontein.”
The El Niño weather phenomenon, which is expected to make its way to South Africa later this spring, could also influence the timing and severity of this year’s allergy season. In South Africa, it is typically associated with hot and dry conditions, but certain regions could receive increased precipitation, along with higher-than-average temperatures leading to enhanced pollen production.
“The underlying influence of climate change makes the country extremely vulnerable to the effects of El Niño,” remarks Prof Peter. “In Europe, El Niño has caused extreme weather events the last few months, where some countries have reported earlier starts and extended pollen seasons. In the UK, record grass counts were recorded.”
SAPNET partners with SA Weather Service
Prof Peter says, in addition to the South African Pollen Network providing allergy sufferers with up-to-date pollen counts, its recent partnership with the South African Weather Service (SAWS) will also give the public and healthcare practitioners access to information related to air pollution, which include carbon monoxide, sulfur dioxide and ozone – all of which are linked to the worsening of allergic diseases.
“As the SAPNET expands, the aim is to provide communities throughout the country with integrated information in real time of what is in the air – spanning the spectrum of air pollution, as well as pollens, even microplastics and other irritants that might be in the air we breathe. We’ll also be working with different communities to further research that is being done on indoor pollutants, which also play a big role, particularly in asthma and asthma exacerbation in children.”
Seasonal allergies affect an estimated 18-million South Africans, which significantly decreases productivity and the quality of day-to-day living among sufferers.
Thanks to funding from Clicks, Thermo Fisher, Glenmark Pharmaceuticals, Cipla and Dr Reddy’s, SAPNET currently covers 10 biomes, including Cape Town, George, Gqeberha-Port Elizabeth, Durban, Johannesburg, Pretoria, Potchefstroom, Bloemfontein and Kimberley.
For up-to-date pollen counts per region, visit pollencount.co.za.