Konstantin Kapinchev

On the possibility of producing true real-time retinal cross-sectional images using a graphics processing unit enhanced master-slave optical coherence tomography system

Abstract. In a previous report, we demonstrated master-slave optical coherence tomography (MS-OCT), an OCT method that does not need resampling of data and can be used to deliver en face images from several depths simultaneously. In a separate report, we have also demonstrated MS-OCT’s capability of producing cross-sectional images of a quality similar to those provided by the traditional Fourier domain (FD) OCT technique, but at a much slower rate. Here, we demonstrate that by taking advantage of the parallel processing capabilities offered by the MS-OCT method, cross-sectional OCT images of the human retina can be produced in real time. We analyze the conditions that ensure a true real-time B-scan imaging operation and demonstrate in vivo real-time images from human fovea and the optic nerve, with resolution and sensitivity comparable to those produced using the traditional FD-based method, however, without the need of data resampling.

Master slave en-face OCT/SLO

Master Slave optical coherence tomography (MS-OCT) is an OCT method that does not require resampling of data and can be used to deliver en-face images from several depths simultaneously. As the MS-OCT method requires important computational resources, the number of multiple depth en-face images that can be produced in real-time is limited. Here, we demonstrate progress in taking advantage of the parallel processing feature of the MS-OCT technology. Harnessing the capabilities of graphics processing units (GPU)s, information from 384 depth positions is acquired in one raster with real time display of up to 40 en-face OCT images. These exhibit comparable resolution and sensitivity to the images produced using the conventional Fourier domain based method. The GPU facilitates versatile real time selection of parameters, such as the depth positions of the 40 images out of the set of 384 depth locations, as well as their axial resolution. In each updated displayed frame, in parallel with the 40 en-face OCT images, a scanning laser ophthalmoscopy (SLO) lookalike image is presented together with two B-scan OCT images oriented along rectangular directions. The thickness of the SLO lookalike image is dynamically determined by the choice of number of en-face OCT images displayed in the frame and the choice of differential axial distance between them.

GPU implementation of cross-correlation for image generation in real time

This paper presents an approach for parallel implementation of cross-correlation using the graphics processing unit (GPU). Cross-correlation is a central digital signal processing (DSP) algorithm with applications in many areas. In many cases in real time systems, a sequential implementation of the cross-correlation creates a performance bottleneck and prevents the systems from reaching the real time criterion. On the other hand, a GPU-based parallel implementation of the cross-correlation offers a solution to this problem. The proposed parallel implementation is integrated in an optical coherence tomography (OCT) system. As a result, the OCT system is able to generate up to 40 en-face images from different depths from semitransparent objects in real time. This number of images provides the necessary information when OCT is used in areas such as ophthalmology, where detailed imagery of the retina, the optic nerve, and other parts of the eye is essential for accurate diagnosis.

Coarse-grained and fine-grained parallel optimization for real-time en-face OCT imaging

This paper presents parallel optimizations in the en-face (C-scan) optical coherence tomography (OCT) display. Compared with the cross-sectional (B-scan) imagery, the production of en-face images is more computationally demanding, due to the increased size of the data handled by the digital signal processing (DSP) algorithms. A sequential implementation of the DSP leads to a limited number of real-time generated en-face images. There are OCT applications, where simultaneous production of large number of en-face images from multiple depths is required, such as real-time diagnostics and monitoring of surgery and ablation. In sequential computing, this requirement leads to a significant increase of the time to process the data and to generate the images. As a result, the processing time exceeds the acquisition time and the image generation is not in real-time. In these cases, not producing en-face images in real-time makes the OCT system ineffective. Parallel optimization of the DSP algorithms provides a solution to this problem. Coarse-grained central processing unit (CPU) based and fine-grained graphics processing unit (GPU) based parallel implementations of the conventional Fourier domain (CFD) OCT method and the Master-Slave Interferometry (MSI) OCT method are studied. In the coarse-grained CPU implementation, each parallel thread processes the whole OCT frame and generates a single en-face image. The corresponding fine-grained GPU implementation launches one parallel thread for every data point from the OCT frame and thus achieves maximum parallelism. The performance and scalability of the CPU-based and GPU-based parallel approaches are analyzed and compared. The quality and the resolution of the images generated by the CFD method and the MSI method are also discussed and compared.

Parallel approaches to integration with applications in optical coherence tomography

This paper presents a comparison between three GPU-based parallel approaches to integration in digital signal processing (DSP). In many cases in practice, the integration is applied on large signals. Its implementation in real-time systems imposes strict limitations on the computational time. In cases where sequential implementation does not meet these limitations, parallel optimization is expected to provide a solution. The discussed parallel approaches are based on the many-core architecture of the GPU. These approaches are employed in generating synthesized scanning laser ophthalmoscopy (SLO) images. These images are based on multiple en-face images, generated by optical coherence tomography (OCT) system. Optimal parallel approaches, which meet the real-time criterion, are identified.

En-face optical coherence tomography revival

Quite recently, we introduced a novel Optical Coherence Tomography (OCT) method, termed as Master Slave OCT (MS-OCT), especially to deliver en-face images. MS-OCT operates like a time domain OCT, selecting signal from a selected depth only while scanning the laser beam across the sample. Time domain OCT allows real time production of an en-face image, although relatively slowly. As a major advance, the Master Slave method allows collection of signals from any number of depths, as required by the user. MS-OCT is an OCT method that does not require resampling of data and can be used to deliver en-face images from several depths simultaneously. However, as the MS-OCT method requires important computational resources, the number of multiple depth en-face images produced in real-time is limited. Here, we demonstrate that taking advantage of the parallel processing feature of the MS-OCT technology by harnessing the capabilities of graphics processing units (GPU)s, information from 384 depth positions is acquired in one raster with real time display of 40 en-face OCT images. These exhibit comparable resolution and sensitivity to the images produced using the traditional Fourier domain based method. The GPU facilitates versatile real time selection of parameters, such as the depth positions of the 40 images out of a set of 384 depth locations, as well as their axial resolution. Here, we present in parallel with the 40 en-face OCT images of a human tooth, a confocal microscopy lookalike image, together with two B-scan OCT images along rectangular directions.

In-vivo, real-time cross-sectional images of retina using a GPU enhanced master slave optical coherence tomography system

In our previous reports we demonstrated a novel Fourier domain optical coherence tomography method, Master Slave optical coherence tomography (MS-OCT), that does not require resampling of data and can deliver en-face images from several depths simultaneously. While ideally suited for delivering information from a selected depth, the MS-OCT has been so far inferior to the conventional FFT based OCT in terms of time of producing cross section images. Here, we demonstrate that by taking advantage of the parallel processing capabilities offered by the MS-OCT method, cross-sectional OCT images of the human retina can be produced in real-time by assembling several T-scans from different depths. We analyze the conditions that ensure a real-time B-scan imaging operation, and demonstrate in-vivo real-time images from human fovea and the optic nerve, of comparable resolution and sensitivity to those produced using the traditional Fourier domain based method.

Master-Slave optical coherence tomography for parallel processing, calibration free and dispersion tolerance operation

We present further improvements on the Master Slave (MS) interferometry method since our first communication [1]. In this paper, we present more data collection and additionally demonstrate an important feature of the MS method, that of tolerance to dispersion. MS interferometry produces the interference of a selected point in depth based on principles of spectral domain (SD) interferometry, but without the need of a Fast Fourier transformation (FFT). The method can be used to directly produce en-face optical coherence tomography (OCT) images but also as a tool to accurately measure distances in low coherence interferometry for sensing applications [1]. In the MS-OCT method, cross-correlation is applied to both methods of SD-OCT, spectrometer based (SP) or swept source (SS) OCT. The channelled spectrum provided by an OCT system is correlated with the signal produced by reading a stored mask. Several such masks can be used simultaneously. The masks operate as adaptive filters. Each mask (filter) determines recognition in the measured channelled spectrum delivered by the interferometer, of the pattern corresponding to each optical path difference to be recognized. The method presents net advantages in comparison with the classical method of producing axial reflectivity profiles by FFT: no need for resampling of data, possibility to tailor the trade-off between depth resolution and sensitivity. Here, using a swept source, the MS method is used to obtain axial reflectivity profiles, which are compared to the axial profiles obtained by calibration of data and FFT. The tolerance to dispersion of the MS method was assumed in [1] but not demonstrated. Here, measurements are performed to demonstrate its axial resolution independence on dispersion.

Real-Time calibration-free C-scan images of the eye fundus using Master Slave swept source optical coherence tomography

Recently, we introduced a novel Optical Coherence Tomography (OCT) method, termed as Master Slave OCT (MS-OCT), specialized for delivering en-face images. This method uses principles of spectral domain interfereometry in two stages. MS-OCT operates like a time domain OCT, selecting only signals from a chosen depth only while scanning the laser beam across the eye. Time domain OCT allows real time production of an en-face image, although relatively slowly. As a major advance, the Master Slave method allows collection of signals from any number of depths, as required by the user. The tremendous advantage in terms of parallel provision of data from numerous depths could not be fully employed by using multi core processors only. The data processing required to generate images at multiple depths simultaneously is not achievable with commodity multicore processors only. We compare here the major improvement in processing and display, brought about by using graphic cards. We demonstrate images obtained with a swept source at 100 kHz (which determines an acquisition time [Ta] for a frame of 200×200 pixels2 of Ta =1.6 s). By the end of the acquired frame being scanned, using our computing capacity, 4 simultaneous en-face images could be created in T = 0.8 s. We demonstrate that by using graphic cards, 32 en-face images can be displayed in Td 0.3 s. Other faster swept source engines can be used with no difference in terms of Td. With 32 images (or more), volumes can be created for 3D display, using en-face images, as opposed to the current technology where volumes are created using cross section OCT images.