Based on the successful development and clinical translation of novel imaging biomarkers in dementia, we are translating our technology into multiple sclerosis (MS) in collaboration with the NMR Research Unit (ION). The aim of this collaboration is to develop novel image analysis pipeline for MS. This work is part of our overall activity to provide a consistent software platform to neuroimaging biomarkers development within the BRC Computational Imaging Infrastructure High Impact Initiative.
The main aspect is focused on brain atrophy quantification and MS lesion filling in order to develop new imaging biomarkers.
Brain atrophy measurements using structural MRI provide an accurate biomarker of MS pathology that is correlated with clinical disability. We have developed a generalised and extended formulation of the Boundary Shift Integral (gBSI). In brief, the developed tool enables, from multiple brain MRI acquisitions of a single patient at multiple time points to, robustly quantify global or regional brain volume change. The proposed process provides a fully automated and highly robust methodology for image analysis, without the need for labour intensive and time consuming human interaction. gBSI has been tested on large dataset and has been shown to out-perform the state-of-the-art methods. It enables to reduce the required sample size to assess group differences, increasing the reliability in a trial of a putative disease-modifying drug.
Figure 1: Diagram representing the GBSI different processing steps for atrophy estimation.
MS lesions influence the process of image analysis, miss-leading to erroneous tissue segmentation or biased morphometric estimates. With the aim of reducing this bias, we have been developing a new multi-modal and multi-time-point lesion filling strategy inspired by computer graphics in-painting techniques for image completion. This technique makes use of a patch-based Non-Local Means algorithm that fills the lesions with the most plausible texture. We have demonstrated that this strategy neglects the lesion-based bias for further image processing such as segmentation, registration and statistical analysis.
Figure 2: Diagram representing the fill lesions technique over a multi-modality image. In short, the proposed method moves the search patch S (in blue), centred at q, within the green area, until it finds the location S(q) that is the most similar to the target patch T (in red), centred in p. Below, we've illustrated the fill lesion technique principle.
All this work has been integrated into our NifTK platform.