School of Mathematics, Statistics & Computer Science

Deep Learning Used to Discover Brain Tumours and Predict Survival

Dr Tirivangani Magadza graduated from the School of Mathematics, Statistics and Computer Science, where his research focused on how deep learning helps detect brain tumours from MRI scans and predict patient survival rates.

Deep learning, a branch of artificial intelligence (AI), enables computers to analyse large amounts of data from MRI scans. AI can quickly and accurately detect tumours, including the tumour core, surrounding swelling and the rapidly growing, aggressive parts.

By using patterns from previous patients (such as age and tumour type), AI can also estimate the survival time for current patients. This technology assists doctors in customising treatments for each patient based on AI-driven insights.

Magadza, a passionate Software Engineer, has always aspired to become an artificial intelligence research and application expert. His journey was challenging, however, as he lacked the financial resources to pursue his dream.

Fortunately, a colleague, Dr Kudakwashe Zvarevashe, informed him that Professor Serestina Viriri from the University of KwaZulu-Natal (UKZN) was seeking motivated PhD students to work on computer vision research. Learning that UKZN offered a fee remission programme, Magadza seized the opportunity to make his dream of researching AI a reality with Viriri as his supervisor.

One of the critical areas of Magadza’s research involved patients diagnosed with glioblastoma multiforme (GBM), the most aggressive type of brain tumour (classified as World Health Organization (WHO) grade IV). Even with intense treatment, including surgery, radiation and chemotherapy, patients with GBM typically have a life expectancy of about 14 months.

Magadza’s research was inspired by the challenges doctors face when accurately identifying and outlining these tumours in MRI scans. Even highly skilled radiologists struggle to outline brain tumours consistently from the same MRI scans, which highlighted the need for a reliable, automated method to improve tumour detection and treatment planning.

This led Magadza to focus his research on developing highly effective and efficient brain tumour segmentation techniques. His goal was to create automated methods to analyse MRI scans, particularly for identifying areas of the brain affected by tumours, like gliomas. After identifying the tumour, Magadza aimed to predict how long a patient might survive.

One of the key aspects of his research was finding ways to make these AI-based solutions less resource-intensive, which is crucial for hospitals and researchers in developing countries where high-tech solutions may be too costly or unavailable. Magadza’s work aimed to make brain tumour diagnosis and survival prediction more accessible by developing methods that use fewer resources while maintaining accuracy.

Looking to the future, Magadza plans to work with organisations to develop AI further and make it accessible to people across Africa, with the goal of uplifting marginalised communities. ‘Together, we can bridge this gap and unlock the full potential of AI,’ he said.

He expressed gratitude to his supervisor Viriri; his wife Addlight; Dr G Shana; Pastor MB Shana; and his parents, family and friends for their unwavering support throughout his journey.

Viriri stated: ‘Dr Magadza has demonstrated tremendous passion throughout his PhD studies, making the journey enjoyable despite the major and minor challenges encountered along the way. He achieved innovative results that were published in top Quartile 1 category journals. Dr Magadza possessed the tenacity to achieve even greater accomplishments in research and has paved the way for prospective postgraduate students to pursue research in the area of Machine Learning for Medical Image Analysis.’

Magadza offered advice to prospective students, saying that while the PhD journey is challenging, the destination provides invaluable experience and personal growth. He summed up his passion for coding with the words: ‘At the end of the day, if I am not coding, well… I am probably still coding!’

Words: Leena Rajpal

Photograph: Sethu Dlamini