Fandom

Scratchpad

Optical imaging/future

< Optical imaging

216,076pages on
this wiki
Add New Page
Discuss this page0 Share

Ad blocker interference detected!


Wikia is a free-to-use site that makes money from advertising. We have a modified experience for viewers using ad blockers

Wikia is not accessible if you’ve made further modifications. Remove the custom ad blocker rule(s) and the page will load as expected.



Future challenges

Experimental techniques

Improvements in source technology are likely to dominate developments in instrumentation. The development of new laser diodes has had a major impact on systems designed for diffuse optical imaging, and further improvements in the technology are anticipated. Higher power will allow increased penetration and faster image acquisition. Significant improvements can be expected from a combination of better theoretical understanding of the optimal combinations of wavelengths for a particular application, and the availability of sources at a wider range of wavelengths.

Molecular imaging is a developing application for diffuse optical imaging techniques, driven partly by the rapid progress which has been made in the development of new dyes and molecular markers which bind to specific molecular targets. Rapid growth in this area is likely to continue both for fundamental scientific research and as mechanisms for evaluating new therapies. Optical imaging is well placed to form an intrinsic role in molecular imaging especially with the availability of novel, specifically targeted, efficient molecular beacons. In particular, quantum dots can act as highly efficient fluorophores whose optical and biochemical characteristics can be finely controlled (Salata 2004, Kim et al. 2004). New femtosecond laser sources and faster detectors (Steinmeyer 2003) may lead to optical imaging systems designed specifically for small animal imaging.

Modelling and reconstruction

The growth in the use of prior information for both modelling and the inverse problem can be expected to continue. Simultaneous recording of anatomical and optical data for both breast and brain imaging will provide knowledge of the internal and external geometry of the tissue, as well as the precise locations of the source and detector fibres, leading to more accurate forward models. If the location of a breast lesion can be identified by anatomical imaging, the functional parameters can then be determined optically. The first steps towards this approach have been taken by Ntziachristos et al. (2002c), Brooksby et al. (2003) and Li et al. (2003). The use of prior information for brain imaging is less well developed, but could be expected to provide even greater improvement in image quality than for breast imaging. An MRI image of the head can provide the location of the non-scattering regions, allowing these to be modelled correctly, possibly by using a coupled radiosity-diffusion model. Furthermore, diffusion tensor imaging can provide information about the direction of the anisotropic scatter of brain tissue. Over the next few years we can expect to see similar advances in the use of prior information in the inverse problem, as theoretical techniques improve along with computer processing power. One promising approach is to use Bayesian techniques to generate not a single image, but a series of images which give the entire posterior probability, which takes into account both measurement and image covariances. An alternative approach might be to address the clinical issue directly and appreciate that it is not always necessary to solve the ill- posed and underdetermined imaging problem. For example, the clinical question may be “is this lesion a tumour?”, or even “is the left side of the neonatal brain different from the right side?”. The answer to such questions may be a single estimate of probability (or a 95% confidence range) and the solution may then become overdetermined and less ill-posed.

The problems of creating good quality finite element meshes which incorporate high resolution anatomical prior information, whilst still being sparse enough to be solved efficiently in the inverse problem, remain to be solved. We can expect to see further developments in meshing techniques such as adaptive mesh refinement to optimise the performance of a finite element mesh. Multigrid methods, which vary the coarseness of the mess as the reconstruction proceeds, have been shown to reduce computation time and improve image quality (Ye et al. 2001) and are likely to become more widely used.

Traditional algorithms for non-linear image reconstruction, such as those discussed in section 3.4, are generally based on iteratively minimising a cost function. This type of approach is slow because the forward problem must be solved repeatedly, and an updated distribution of optical properties must be found at each step. Recently, the first direct non-linear reconstruction techniques have been implemented in EIT (Siltanen et al. 2000) based on a proof by Nachman (1996) in which the non-linear reconstruction problem was transformed to two linear integral equations. Mueller et al. (2002) applied the method to a realistic simulation, and Isaacson et al. (2004) reconstructed data acquired on an anatomically realistic phantom. In both cases, the method gave qualitatively and quantitatively accurate results. In its current form, there are disadvantages: it has not yet been shown how to extent the method to optical tomography, how it can be extended to 3D, or how prior information can be included. Other approaches to direct reconstruction have been suggested in EIT (Lionheart 2004) and it is likely that these potentially important new developments will eventually be applied to diffuse optical reconstruction.

Clinical applications

The use of NIR techniques to image haemodynamic changes is now well established. In neurophysiological terms, however, the electrical activity of neurones is of primary interest and the haemodynamic response is a secondary effect. However, optical techniques have already been used to examine neuronal activity in cell cultures (Cohen 1973, Lazebnik et al. 2003) and in the exposed brain (Rector et al. 2001). This small, fast signal has been recorded noninvasively in humans by Gratton et al. (1997) and Gratton and Fabiani (2003) using a frequency-domain system (ISS Inc, USA) and was shown to correlate temporally with EEG and spatially with the haemodynamic response imaged by fMRI. The signal has been observed in measurements of both phase and amplitude (Wolf et al. 2003). More recently, Franceschini and Boas (2004) used a CW system to record fast neuronal signals from approximately 60% of subjects following 1000 averaged active or passive motor repetitions (see Figure 7).

The nature of the relationship between neuronal activity and the associated haemodynamic response, known as neurovascular coupling, is not well understood. This can lead to uncertainties in the interpretation of brain imaging techniques which are sensitive to haemodynamic effects. The ability of optical methods to image both phenomena simultaneously could provide a unique system for the examination of the coupling between the two effects and thereby lead to new insights into brain physiology (Lauritzen and Gold 2003). This could have particular advantages for the assessment of brain development in infants and for the investigation of certain neurological conditions. For example, epilepsy may be initiated by pathological neuronal activity in deeper brain structures which later propagates to the cortex where it manifests as an increase in blood flow. Optical tomography may be able to localise the initial focus of epilepsy, providing more accurate diagnosis and leading to improved therapy.

One of the most challenging future roles for diffuse optical imaging is 3D imaging of the newborn infant brain, which faces unique problems due to the vulnerability of the subject. In order that optical tomography of the infant brain can have the greatest impact on mortality and the incidence of permanent brain injury, the task over the next few years is to develop a technique which can be employed rapidly and easily at the bedside, within hours of birth. Means must also be developed for monitoring the brain over periods of hours or days in order to assess the response to new forms of treatment. In future, optical tomography may also aid the study of the healthy developing brain via the increased scatter produced by myelin, as it develops around neurones.

Optical mammography has been widely evaluated as a clinical tool. Large scale clinical trials involving 300 patients have been carried out as part of the EU funded OPTIMAMM project (section 4.3) and have identified 80-85% of tumours (Grosenick et al. 2004a, Taroni et al. 2004b). A new multi-centre network funded by the US National Cancer Institute is likely to lead to significant developments by bringing together groups with different areas of expertise (NTROI: www.bli.uci.edu/NTROI/indexmain.html). It is likely that the clinical utility of optical mammography will increase as more wavelengths provide improved distinction between chromophores, and full tomographic reconstruction improves spatial discrimination. Furthermore, the use of optical mammography to provide functional information as an adjunct to anatomical imaging techniques is still in its infancy. The application of molecular imaging techniques to optical mammography could lead to particularly exciting clinical advances (Ntziachristos et al. 2002a).

One aspect of optical imaging which is common to all applications is the development of improved ways to attach sources and detectors to the body. A range of approaches has been developed, from rigid arrays for mammography (Dehghani et al. 2003b) to moulded pads for brain imaging (Hebden et al. 2002, Koizumi et al. 2003), and even noncontact imaging (Schulz et al. 2003). Unwanted physiological effects can be the dominant source of error in optical imaging and so optimising the location of the source and detector fibres and minimising the coupling effects can lead to significant improvements in image quality (Strangman et al. 2003, Franceschini and Boas 2004).

Also on Fandom

Random wikia