Contactless methods for fast micro-volume visualization based on low-coherence interferometry
Information about an object structure is needed in many areas of science and modern technology. For a long time, the only available method of visualization of such information was being represented in the form of plain (2D) image containing information about light intensity distribution at the object (scene) plane. With appearance of the methods to extend this information (for example, methods for determining the distance to each point, thermography methods, etc.), there became necessary to visualize it. In the framework of the two-dimensional approach, this is possible, for example, using pseudo-colors or an additional 2D image, containing, for example, information about the distance to each point of scene. Development of computational means allowed to implement algorithms for visualization of pseudo 3D objects (surfaces) on flat screen. Utilization of modern graphic processors capabilities allows to visualize "full" 3D images (brightness at each voxel), but the methods for obtaining information for this are very time consuming. For a long time, visualization of such information was not highly demanded, especially in tasks involving subjective (expert) analysis of images. For example, in medicine, an ultrasound scan is available for medical doctor at any time showing a cross-sectional image in depth (B-scan) of an object under study, and it is not necessary to obtain simultaneously full information about three-dimensional structure due to complexity of its cumulative human analysis. Nevertheless, the methods of instantaneous visualization of a 3D structure are relevant to the problems, when, for instance, structure of an object changes in time. 3D video recording techniques are needed also to improve stability and reliability of automated data analysis when diagnosing an object condition.
Optical coherence tomography (OCT) is one of the promising and popular methods of obtaining information on the internal microstructure of objects in medicine and in a number of technical applications. In majority of commercial systems of spectral OCT, only one B-scan is usually displayed (as in the ultrasonic study), as a result of which the possibility of simultaneous visualization of 3D structure of object is lost. The use of sources with tunable wavelengths and 2D photosensitive arrays (video cameras) allows potentially working in full-field mode and displaying a 3D object microstructure, but the performance of modern photodetectors is not sufficient for rapid volume visualization with acceptable resolution and large field of view. For instance, in the group of J. Isatt (Sci. Rep., 2016, 6: 31689) such a system made it possible to visualize a volume in the task of in vivo visualization during ophthalmic operation with frequency approximately 1 Hz provided by high-performance hardware data processing. At the same time, the cost of systems with tunable wavelength remains very high. In the time-domain OCT systems based on low-coherence interferometer and utilizing scanning into interior of object depth, rather than in vertical cross-section, the algorithms for data processing in off-line mode are mainly used that also makes difficult dynamic volume visualization.
In the scientific group of International Scientific-Technical Center for Computational Optics, Photonics and Imaging of ITMO University, active work is being conducted for several years to create algorithms for dynamic data processing in time-domain OCT systems, as well as the necessary equipment and software is available for creation of such systems. Up to now, algorithms for dynamic analysis of each A-scan based on stochastic filters of the Kalman type have been developed and investigated. However, due to random nature of objects under study and, as a consequence, the recorded data, the result of processing neighboring A-scans may differ, so the result of volume visualization can be unsatisfactory. Within the framework of the proposed project, it is supposed to develop and to study the methods of dynamic data processing in time-domain OCT involving joint processing of a set of A-scans based on methods of multidimensional stochastic filtering. It is also planned to develop a laboratory set-up of OCT system and to test the software that would implement efficient algorithms to allow visualizing object micro structure with frequency not lower than 1 Hz (as in comparable spectral OCT systems), while providing high resolution (up to several times higher than the spectral OCT provides) with significantly lower cost of the device (in comparison with the ones with tunable wavelength light source). It should be noted that in the proposed system, the frequency of "3D video" at 1 Hz, as preliminary estimates determine, is limited by the capacity of modern data transmission interfaces from the video camera. This frequency can be provided by using video cameras with data transmission speed of about 12 MVox/s. Taking this into account, it is proposed to develop a system with variable field of view that allows to perform an express study of a micro-volume with limited lateral resolution, but with large field of view. Then a study of areas of interest with high lateral resolution in a small field of view is performed. At the same time the capabilities of modern graphics processors will be used for data processing as well as the algorithms based on parallel computing. Scientific contribution of the work consists in creation of computational methods and algorithms to ensure option of one-stage visualization of entire volume in dynamic mode, replacing the current technological solutions (in particular, very expensive radiation sources) and providing, nevertheless, functional characteristics required in optical tomography systems.