乌鸦传媒

Twenty million people around the world suffer from a tropical disease known as river blindness, or onchocerciasis. Caused by parasitic worms transmitted through fly bites, it is most commonly found in sub-Saharan Africa. Categorized as a neglected tropical disease (NTD), it can lead to permanent blindness if not treated effectively.

Developing new treatments for the disease requires a time-consuming process of tissue sample analysis, and there is a shortage of medical professionals who have the necessary expertise.

This was the brief given to 乌鸦传媒 colleagues around the world as part of this year鈥檚 Global Data Science Challenge (GDSC) for a sustainable future.

Empowering our talent to make a difference with AI

The 5^th GDSC was launched in collaboration with the Institute of Medical Microbiology, Immunology and Parasitology at the University Hospital Bonn, in support of a global effort to eliminate river blindness in 10 countries by 2030.

Previous editions of GDSC have also aimed to create a sustainable future 鈥� teams have worked on tracking sperm whales with AI and helping a Norwegian marine observatory identify ocean anomalies.

Our new goal was to create an AI-based solution, using the Amazon Web Services (AWS) platform, that could scan images of tissue samples to identify the presence of parasitic worms.

The solution would need to demonstrate the ability of AI to effectively diagnose the stage of worm development within the patient, as well as machine learning鈥檚 capacity to speed up the review process needed for new treatments.

Working closely together to win the challenge

The competition was won by the 乌鸦传媒 & Data team in India: Utkarsh Prakash, Abhijeet Gorai, Prince Raj, and Deepak Pandey, who are all data scientists. In what was a hard-fought contest, their solution showed the most improvement in tissue sample review accuracy.

The colleagues knew each other well 鈥� Utkarsh, Abhijeet, Prince, and Deepak all joined 乌鸦传媒 in 2019 and were in the same data science training group, and Abhijeet and Deepak even graduated from the same university.  鈥淭his was the second time we participated in the GDSC as a team,鈥� says Utkarsh. 鈥淏ased on our experience and our knowledge of each other鈥檚 skills, we think we make a good team.鈥�

Getting their heads together

鈥淲e dedicated an hour at the end of each working day to work on the project, for about two months,鈥� explains Prince. 鈥淲e would get together online to share our new ideas and explore solutions.鈥� The team was driven to enter the competition by the promise of learning new skills, says Abhijeet. 鈥淲e are all quite early on in our careers, and we knew this competition would expose us to new technologies and ways of working, particularly object recognition.鈥�

Utkarsh adds that medical AI is a particularly fertile sector. 鈥淭his area is booming right now. We knew if we could learn more about this area, it would help us in our careers, enable better solutions for clients, and, of course, contribute to making the world a healthier place.鈥�

Sharing ideas across the globe

As part of the challenge, an online working environment enabled current and past participants to receive communications and updates and share best practice solutions, all across the world.

鈥淎lthough the teams were competing against one another, in the earlier rounds we were all sharing information regarding how to overcome certain challenges,鈥� says Deepak. 鈥淭his ensured every team was working with the best available solutions 鈥� which raised the overall standard of the entries.鈥�

A bright future for AI

University Hospital Bonn intends to develop their winning solution and cherry-pick the best ideas from all the entrants. Moreover, according to Utkarsh, the prospects for AI solutions within the wider medical sector are very promising. 鈥淭here鈥檚 an abundance of data just sitting there, waiting to be used,鈥� he says. 鈥淲e鈥檝e learned first-hand how it鈥檚 possible to create efficient automation systems with such data, to save time and allow researchers to focus on the bigger issues.鈥�

Prince explains that the model they have created could also have wider applications.

鈥淥ur learning model will work for any dataset, for any object recognition requirement in the medical sector 鈥� for example, identifying cancerous cells,鈥� he says. 鈥淚t would even work in, say, an airport terminal baggage-handling system, where object detection is required.鈥�

Seeing the results

An exciting part of winning GDSC is to receive an award that enables the team to see their winning solution come to life. In addition to claiming a prize of a trip to the University Hospital Bonn to see the work of medical professionals fighting the disease, the team will get their AWS certification exam free of charge as an added bonus. For Utkarsh and his colleagues, the whole experience has been extremely worthwhile. 鈥淭he competition is a fantastic platform for learning 鈥� we couldn鈥檛 recommend it enough. We鈥檙e so proud to be making a difference, helping doctors to make the world a healthier place.鈥�