Does AI offer a painless diagnosis for brain tumour patients? Can we at least automate the process of finding tumours in the human brain? In a review paper published recently in the journal Digital Health, this is what Professor Jayendra M Bhalodiya, Assistant Professor, School of Engineering and Applied Science, has established. Having reviewed more than 550 research papers with collaborators from the Institute of Digital Healthcare, WMG, at The University of Warwick, UK, Professor Bhalodiya highlights that artificial intelligence methods to diagnose brain tumours are gaining prominence, promising to eliminate painful biopsies forever.
Interpreting his observations, Professor Bhalodiya said, "We need better AI solutions and data to diagnose brain tumours. This review is just the beginning of my research motivated by the need to reduce the number of biopsy tests or avoid them completely, if possible, so that patients won't have to suffer much pain."
The brain tumour segmentation studies were conducted over six years, from 2015 to 2020. The researchers reviewed segmentation techniques using T1-weighted, T2-weighted, gadolinium-enhanced T1-weighted, fluid-attenuated inversion recovery, diffusion-weighted and perfusion-weighted magnetic resonance imaging (MRI) sequences. They assessed physics or mathematics-based methods, deep learning methods, and software-based or semi-automatic methods, as applied to MRI techniques. Notably, they synthesised each method as per the utilised MRI sequences, study population, technical approach (such as deep learning) and performance score measures (such as Dice score).
However, quite understandably, there are some critical requirements before implementing AI as an alternative to manual segmentation or biopsy. "We need comprehensive solutions covering multiple types of MRI in the method development, validation and testing; so that doctors and patients can trust them," said Professor Bhalodiya. If validated, not just patients but also radiographers stand to benefit.
The researchers are involved in ongoing research about identifying MRI image biomarkers in brain tumour patients.
For more details on the systematic review, click on Digital Health