nEUROPt: Non-invasive imaging of brain function and disease by pulsed near infrared light

This project is funded by the Seventh Framework Programme Health (Grant agreement no. 201076) as a collaborative project between several European research centres. Home page: http://www.neuropt.eu/

Project objectives:

Development of novel techniques for optical imaging of the brain, yielding improved spatial resolution, widespread coverage of the head, optimum selectivity to signals from the brain, and absolute quantification of physiological data. This work will be accompanied by parallel developments in modelling of pulsed near-infrared light propagation in the brain. By solving the inverse problem, the improvements in modelling will lead to better lateral and depth resolution, localisation, and quantification of optical properties within various tissue compartments of the head and brain.

Development of new clinical prototype time-domain systems for diffuse optical imaging of the brain, adapted to the needs of specific clinical applications in neurological assessment.

Characterisation of performance of instruments together with basic methods of data analysis. Definition and implementation of standardised protocols for quality assessment.

Assessment of the diagnostic value of time-domain brain imaging by clinical pilot studies which address several major neurological pathologies.

UCL-CS involvement

The main contribution of UCL-CS to the project is in modelling of the forward problem and the improvement of nonlinear reconstruction methods. Modelling of the forward problem has been developed based on two particular physical descriptions: the general radiative transfer equation (RTE) and its simplest approximation, the diffusion equation (DE). The aim of this project is to further develop a hybrid RTE/diffusion model that combines the computational efficiency of the diffusion model with the accuracy of the RTE model in low-scattering regions. A key development are multiscale models with adaptive resolution which increase the accuracy and computational effort only locally, using simpler and coarser models for parts of the domain far from sources and detectors.

Numerical approaches to the forward problem such as the finite element method (FEM) can be applied to complex geometries, but can be time-consuming, in particular where the temporal distribution must be calculated explicitly. We are proposing a fast method for estimating the temporal profile from a small number of moments of the distribution, which can be calculated efficiently.

Nonlinear reconstruction methods are being improved by utilisation of dimension reduction methods based on prior knowledge or region identification, or on implicit or explicit region boundary shapes which are potentially faster and more robust than traditional model-fitting procedures. We are exploiting new developments in approximation error theory, meshless methods, Krylov-Newton solution methods, and parallel architecture to provide efficient, robust and fast reconstruction methods.

In the framework of this project we are investigating three categories of priors and regularisation: generic priors which are based on the expected local image statistics without explicit knowledge of the particular structure, anatomical priors which take explicit account of known anatomical features, and spectral priors which explot the expected spectral characteristics of the chromophores and the scattering coefficient present in tissue. Furthermore, data from other imaging modalities can act as priors for optical reconstruction. In general, determination of the optical properties of brain can be improved if a-priori knowledge of the optical properties of other tissues involved (scalp, bone) is available.