Xiqi Li

Morphometric measurement of Schlemm’s canal in normal human eye using anterior segment swept source optical coherence tomography

We have used anterior segment swept source optical coherence tomography to measure Schlemms canal (SC) morphometric values in the living human eye. Fifty healthy volunteers with 100 normal eyes were measured in the nasal and temporal side. Comparison with the published SC morphometric values of histologic sections proves the reliability of our results. The statistical results show that there are no significant differences between nasal and temporal SC with respect to their diameter, perimeter, and area in our study (diameter: t=0.122, p=0.903; perimeter: t=-0.003, p=0.998; area: t=-1.169, p=0.244); further, no significant differences in SC morphometric values are found between oculus sinister and oculus dexter (diameter: t=0.943, p=0.35; perimeter: t=1.346, p=0.18; area: t=1.501, p=0.135).

Comparison of Schlemm’s canal’s biological parameters in primary open-angle glaucoma and normal human eyes with swept source optical

Abstract. Thirty-seven normal and primary open angle glaucoma (POAG) subjects were noninvasively imaged by a tailor-made real-time anterior segment swept source optical coherence tomography (SS-OCT) to demonstrate the differences of the Schlemm’s canal (SC) between POAG and normal eyes. After the cross-section images of the anterior chamber angle were acquired by SS-OCT, SC was confirmed by two independent masked observers and the average area, long diameter, and perimeter of the SC were measured. In normal subjects the circumference, area, and long diameter is 580.34±87.81  μm, 8023.89±1486.10  μm2, and 272.83±49.39  μm, respectively, and these parameters were 393.25±98.04  μm, 3941.50±1210.69  μm2, and 190.91±46.47  μm in the POAG subjects. The area of SC in the normal ones was significantly larger than that in POAG eyes (p<0.001), so as the long diameter and the perimeter (p<0.001; p<0.001).

Time-domain interpolation for Fourier-domain optical coherence tomography.

The imaging speed and quality of a Fourier-domain optical-coherence-tomography (FD-OCT) technique is largely limited by the resampling process. A time-domain interpolation approach based on zero padding is proposed that gets a close fall-off but much-reduced imaging time in FD-OCT data processing as compared with the common zero-padding interpolation method. The experimental results obtained with an FD-OCT system using a 2048 pixel line scan camera showed that the fall-off was improved by ~2 dB in deep z position and that the processing time was reduced to 468 ms per 400 axial scan lines with a Pentium Dual E2140 computer; that is, a more than 95% reduction compared with the conventional zero-padding approach.

High-speed optical coherence tomography signal processing on GPU

The signal processing speed of spectral domain optical coherence tomography (SD-OCT) has become a bottleneck in many medical applications. Recently, a time-domain interpolation method was proposed. This method not only gets a better signal-to noise ratio (SNR) but also gets a faster signal processing time for the SD-OCT than the widely used zero-padding interpolation method. Furthermore, the re-sampled data is obtained by convoluting the acquired data and the coefficients in time domain. Thus, a lot of interpolations can be performed concurrently. So, this interpolation method is suitable for parallel computing. An ultra-high optical coherence tomography signal processing can be realized by using graphics processing unit (GPU) with computer unified device architecture (CUDA). This paper will introduce the signal processing steps of SD-OCT on GPU. An experiment is performed to acquire a frame SD-OCT data (400A-lines×2048 pixel per A-line) and real-time processed the data on GPU. The results show that it can be finished in 6.208 milliseconds, which is 37 times faster than that on Central Processing Unit (CPU).

Extraction of ultra-high frequency retinal motions with a line scanning quasi-confocal ophthalmoscope

Ultra-high frequency motions of the retina severely affect the stabilization of images and result in significant imaging distortions. In this paper, a high speed line scanning quasi-confocal ophthalmoscope (LSO) capable of 160 frames per second was devised for generating stable undistorted retinal images. This technique resulted in minimal intra-frame motions, and a strip-based cross correlation algorithm with sub-pixel resolution was applied to extract retinal motions. Three retinal motion components at rates of up to 1600 Hz were clearly distinguished and extracted accurately for the first time using ophthalmic imaging methods. This was especially apparent for the fastest tremor and microsaccade movements that cannot be estimated from previously reported ophthalmic imaging instruments. Furthermore, these results were consistent with retinal motion characteristics obtained with optical lever methods, validating this technique. Actually, the LSO system has great potential to extract retinal motions, and some other tracking systems may be adopted to correct retinal motions in ophthalmic imaging modalities.

TIME-DOMAIN INTERPOLATION ON GRAPHICS PROCESSING UNIT

The signal processing speed of spectral domain optical coherence tomography (SD-OCT) has become a bottleneck in a lot of medical applications. Recently, a time-domain interpolation method was proposed. This method can get better signal-to-noise ratio (SNR) but much-reduced signal processing time in SD-OCT data processing as compared with the commonly used zero-padding interpolation method. Additionally, the resampled data can be obtained by a few data and coefficients in the cutoff window. Thus, a lot of interpolations can be performed simultaneously. So, this interpolation method is suitable for parallel computing. By using graphics processing unit (GPU) and the compute unified device architecture (CUDA) program model, time-domain interpolation can be accelerated significantly. The computing capability can be achieved more than 250,000 A-lines, 200,000 A-lines, and 160,000 A-lines in a second for 2,048 pixel OCT when the cutoff length is L = 11, L = 21, and L = 31, respectively. A frame SD-OCT data (400A-lines x 2,048 pixel per line) is acquired and processed on GPU in real time. The results show that signal processing time of SD-OCT can be finished in 6.223 ms when the cutoff length L = 21, which is much faster than that on central processing unit (CPU). Real-time signal processing of acquired data can be realized.

Microscope-integrated optical coherence tomography for image-aided positioning of glaucoma surgery

Abstract. Most glaucoma surgeries involve creating new aqueous outflow pathways with the use of a small surgical instrument. This article reported a microscope-integrated, real-time, high-speed, swept-source optical coherence tomography system (SS-OCT) with a 1310-nm light source for glaucoma surgery. A special mechanism was designed to produce an adjustable system suitable for use in surgery. A two-graphic processing unit architecture was used to speed up the data processing and real-time volumetric rendering. The position of the surgical instrument can be monitored and measured using the microscope and a grid-inserted image of the SS-OCT. Finally, experiments were simulated to assess the effectiveness of this integrated system. Experimental results show that this system is a suitable positioning tool for glaucoma surgery.

Acceleration of optical coherence tomography signal processing by multi-graphics processing units

A multi-GPU system designed for high-speed, real-time signal processing of optical coherence tomography (OCT) is described herein. For the OCT data sampled in linear wave numbers, the maximum processing rates reached 2.95 MHz for 1024-OCT and 1.96 MHz for 2048-OCT. Data sampled using linear wavelengths were re-sampled using a time-domain interpolation method and zero-padding interpolation method to improve image quality. The maximum processing rates for 1024-OCT reached 2.16 MHz for the time-domain method and 1.26 MHz for the zero-padding method. The maximum processing rates for 2048-OCT reached 1.58 MHz, and 0.68 MHz, respectively. This method is capable of high-speed, real-time processing for OCT systems.

Centroid offset estimation in the Fourier domain for a highly sensitive Shack-Hartmann wavefront sensor

The estimation of the centroid offset can have an effect on the accuracy of wavefront measurements conducted by highly sensitive Shack-Hartmann wavefront sensors. In this paper, a novel offset estimation algorithm processed in the Fourier domain is proposed. This method can be used to process the offset estimation in the Fourier domain and is efficient in noise suppression. The principle of the algorithm is described in detail. Comparisons between the technique and two other widely used algorithms, the best-threshold center of gravity algorithm and the correlation algorithm, are performed theoretically using numerical simulation and experimentally using a Shack-Hartmann wavefront sensor. The results show that the proposed offset estimation algorithm is unbiased, as robust as the correlation algorithm, as fast as the best-threshold center of gravity algorithm, and achieves a good balance between precision and speed.

HIGH-SPEED SPECTRAL DOMAIN OPTICAL COHERENCE TOMOGRAPHY SIGNAL PROCESSING WITH TIME-DOMAIN INTERPOLATION USING GRAPHICS PROCESSING UNIT

Sensitivity and data processing speed are important in spectral domain Optical Coherence Tomography (SD-OCT) system. To get a higher sensitivity, zero-padding interpolation together with linear interpolation is commonly used to re-sample the interference data in SD-OCT, which limits the data processing speed. Recently, a time-domain interpolation for SD-OCT was proposed. By eliminating the huge Fast Fourier Transform Algorithm (FFT) operations, the operation number of the time-domain interpolation is much less than that of the zero-padding interpolation. In this paper, a numerical simulation is performed to evaluate the computational complexity and the interpolation accuracy. More than six times acceleration is obtained. At the same time, the normalized mean square error (NMSE) results show that the time-domain interpolation method with cut-off length L = 21 and L = 31 can improve about 1.7 dB and 2.1 dB when the distance mismatch is 2.4 mm than that of zero-padding interpolation method with padding times M = 4, respectively. Furthermore, this method can be applied the parallel arithmetic processing because only the data in the cut-off window is processed. By using Graphics Processing Unit (GPU) with compute unified device architecture (CUDA) program model, a frame (400 A-lines × 2048 pixels × 12 bits) data can be processed in 6 ms and the processing capability can be achieved 164,000 line/s for 1024-OCT and 71,000 line/s for 2048-OCT when the cut-off length is 21. Thus, a high-sensitivity and ultra-high data processing SD-OCT is realized.