Indore (Madhya Pradesh): A study at the intersection of biology, statistics and medical imaging by IIM Indore faculty member Prof Sujoy Mukhoti has resulted in the development of a new probability model called the ‘stomped normal (SN) distribution’ for improving the interpretation of MRI brain scans. Traditional MRI analysis relies on unimodal data, which can be limited in its accuracy.
However, the SN distribution model considers a wider range of data, particularly in those areas of the brain where the data distribution is flat. Mukhoti co-investigated a Royal Society-funded project aimed at enhancing MRI brain imaging techniques through advanced probability modelling. The study, which explores how probability models can better interpret MRI scans of the brain, is poised to offer more accurate diagnoses and treatment planning for medical professionals.
“MRI brain scans are crucial diagnostic tools capturing the brain’s internal structure by using magnetic fields and radio waves. However, traditional MRI analysis relies heavily on unimodal data, which focuses on the most common data points observed in a brain region. Such an approach, though effective, can limit the nuance captured in the image data”, Mukhoti said.
He and his team introduced a new class of probability distributions known as ‘SN distribution’, which is capable of handling flatter, more complex data sets than the widely used normal distribution.
“Unlike the unimodal approach, which relies on a single peak of data, the stomped normal distribution allows for a more detailed analysis by considering a wider range of values within MRI data, particularly in brain regions where the data distributions are flat or ambiguous”, the researcher explained.
“The study explores how statistical modelling, specifically new types of probability distributions, can improve our understanding and interpretation of brain MRIs. By incorporating a larger cluster of data, we can better capture the nuances of the brain’s grey matter, offering more precision in medical diagnoses”, Mukhoti said.
The new approach provides more precise and nuanced readings of brain structures, such as grey matter, leading to better accuracy in medical diagnoses and treatment planning. The study demonstrated that the stomped normal distribution outperformed traditional models, especially in complex, flat-top data distributions. The study employed both real and simulated MRI data, demonstrating the robustness of the stomped normal model compared to traditional methods.
