Motion compensation in fiber-optic confocal microscopy
Confocal microscopy is a widely used high-resolution optical imaging technique that provides both high lateral and axial resolution. Importantly, it has been extremely valuable for biomedical applications, such as deep brain and 3D imaging of biological specimens.1 Endoscopic confocal microscopes that use flexible coherent fiber bundles—consisting of tens of thousands of fiber channels—have also been implemented for high-resolution imaging.2, 3 However, in vivo endoscopic confocal images suffer loss of focus because of movement in the target, caused by, for example, breathing and cardiac activities.4 To correct these intra- and inter-frame distortions, motion compensation is crucial. Here, we incorporate a common-path Fourier domain optical coherence tomography (CP-FDOCT) distance sensor to compensate for involuntary movement.
Motion of the sample can be generally divided into either axial or lateral directions. Lateral motion causes deformation in the acquired image, which can be corrected by a software approach.5 However, successful correction using software requires the image to be clearly in focus. Thus, axial motion compensation is a prerequisite for further image correction for fiber-optic confocal microscopy. While the motion effects can be reduced by increasing the imaging speed of the system, rapid imaging cannot always be achieved because of low signal intensity, which requires longer signal integration time.
To solve axial motion issues and to track motion, we used a fiber-optic distance sensor based on CP-FDOCT. Using the sensor, we can measure the distance between the confocal imaging probe and target (see Figure 1). We can then use this measurement to control the position of the imaging probe to ensure that the target is in focus. CP-FDOCT has recently been shown to precisely sense distance and allow for compact instrument integration.6, 7 In our system, we secured an optical coherence tomography (OCT) probe, which uses a single-mode optical fiber, along the fiber bundle confocal imaging probe attached to the shaft of a high-speed linear motor. We used a computer to monitor the distance and deliver control commands to the linear motor through a motor driver. We predetermined the ideal imaging distance (D), which is the functional distance of the fiber bundle probe where the acquired image is sharpest and clearest. While our CP-FDOCT system measures the actual distance (d), the distance error (e)—defined as d−D—is analyzed by the program. If the error is within a 2 pixel distance, no action is required. However, if e>2 pixels, the motor moves the probe at a corresponding velocity to adjust so e<2 pixels.
To test the concept, we built a fiber-optic confocal microscope using a fiber bundle with 10,000 fiber cores. In this set-up, the imaging plain was oversampled by 200×200 pixels (460×460μm). We attached a gradient index lens imaging system at the end of the fiber bundle. The data card was set at a sampling rate of 40,000/s, which set the imaging frame rate to 1fps. The axial resolution of the confocal system was 40μm. We used the NBS 1963A resolution target as a sample because of its sharp image contrast and its small numerical markers that fill the field of view of the probe. We placed the sample on a translation stage, which was moved back and forth relative to the focal plane to simulate axial motion. Our CP-FDOCT system exhibited an axial resolution of 3.6μm in air and 2.8μm in water. Additionally, the linear motor specifications include a 35mm travel range, 20mm/s maximum speed, and less than 1nm resolution depending on different control modes. Specifically, the motor can move in full, shorter, or partial steps to give position resolution in the nanometer range, which gives us greater control in motion compensation. We calibrated the peak position accuracy at 1.6μm/pixel in air. We could achieve higher position accuracy by zero-padding the OCT spectrum. Finally, we achieved a motion compensation rate of 840Hz during the experiment.
Figure 2 shows sequential confocal images with and without motion compensation. We can clearly see that the CP-FDOCT-based motion compensation system can track the focal plane effectively to provide clear, in-focus images. The average amplitude of motion added to the sample stage was 60μm at a frequency of 0.3Hz. The average speed of sample motion was 72μm/s. Additionally, when compensation was enabled, the focusing error was within 5μm. We obtained the theoretical maximum sample motion speed (Vsample), defined as: 
where Tloop is the control loop constant time, Tsampling is the acquisition time for each pixel, Vmotor is the motor speed, and n stands for the number of pixels requiring compensation. If we want every pixel to be compensated and to drive the motor at full speed, our present system can compensate with a motion speed up to 411μm/s. Notably, when the sampling rate is at 40,000/s, the sample displacement during 1 pixel signal acquisition time is only 0.5μm. That is, the distortion for such a small defocus is negligible.
In summary, we have developed an axial motion compensation system for fiber-optic confocal microscopy based on a CP-FDOCT distance sensor. Our system is able to compensate axial motion with amplitudes up to 60μm at an average speed of 72μm/s with a compensation rate of 840Hz. The sample focus error was kept within 5μm. Calculations shows that the upper limit of motion speed that our system can compensate is 411μm/s. We are currently working on optimizing a peak detection algorithm, which is a control method to reduce the error offset and increase the system speed.
Yong Huang has been a PhD candidate since 2009. He is a recipient of the Chinese Scholarship Council Fellowship.
Jin Kang is a professor and the department chair whose research areas include fiber optic devices, OCT imaging and sensing, and biophotonics. He is a fellow of the Optical Society of America.
