Image analytics involves the use of imaging data to derive quantitative measures of physiological parameters.
MRI is only able to directly measure properties of water in different tissues of the body. To link these properties to physiology we need a quantitative forward model of the data which predicts imaging outputs from parameters of interest.
The first step in modelling involves preprocessing of the imaging data to correct for distortions in the imaging process, align separately acquired imaging data, and extract structural information, for example segmentation of the image by organ or tissue type.
A parameterised model specific to the imaging acquisition is then fitted to the data typically using a Bayesian approach in which prior knowledge of parameter ranges can be incorporated and uncertainties in the fitted parameter values can be quantified. Fitted parameters can be used to derive structural averages in regions of interest or maps of functional connectivity.
The aim is to derive through this analysis a set of quantitative measures known as imaging derived phenotypes (IDPs) for the imaging session. IDPs can be linked with other data such as clinical outcomes or genetic sequencing to gain insights into the causes and progression of disease. IDPs can also be used as training inputs for machine learning models of organ health, for example 'Brain age'.