Accurate real-time depth control for CP-SSOCT distal sensor based handheld microsurgery tools

Abstract

This paper presents a novel intuitive targeting and tracking scheme that utilizes a common-path swept source optical coherence tomography (CP-SSOCT) distal sensor integrated handheld microsurgical tool. To achieve micron-order precision control, a reliable and accurate OCT distal sensing method is required; simultaneously, a prediction algorithm is necessary to compensate for the system delay associated with the computational, mechanical and electronic latencies. Due to the multi-layered structure of retina, it is necessary to develop effective surface detection methods rather than simple peak detection. To achieve this, a shifted cross-correlation method is applied for surface detection in order to increase robustness and accuracy in distal sensing. A predictor based on Kalman filter was implemented for more precise motion compensation. The performance was first evaluated using an established dry phantom consisting of stacked cellophane tape. This was followed by evaluation in an ex-vivo bovine retina model to assess system accuracy and precision. The results demonstrate highly accurate depth targeting with less than 5 μm RMSE depth locking.

Keywords: (060.2370) Fiber optics sensors; (150.5758) Robotic and machine control; (170.4460) Ophthalmic optics and devices; (170.4500) Optical coherence tomography.

Fig. 1

Three consecutive A-scan data of 20-stacked layers of adherent cellophane tape near the DC line (multi-layered phantom sample); the upper right inserted graph is a partial magnified graph of area in dotted red square in lower entire graph.

Fig. 2

a) Active depth targeting and locking microsurgery system, b) CAD cross-sectional image of handheld tool (NTC: needle-to-tube connector, PM: piezo-motor, OF: optical fiber).

Fig. 3

a) A-scan data of ex-vivo open bovine retina, b) three consecutive A-scan data of stacked layers of adherent cellophane tape, c) data processing flowchart consisting of 1) OCT data process, 2) Surface detection, and 3) Motor control.

Fig. 4

Dry phantom experiment: a) evaluation test for the surface detection algorithm with the auxiliary shifted cross-correlation in non-depth-locking freehand, b) evaluation test for the depth-locking with the predictor based on Kalman filter: the upper inserted graph is a magnified graph of area in dotted yellow trapezoid in the entire graph.

Fig. 5

Bovine retina experiment: a) bovine retina after removing cornea, lens, and vitreous humor, b) snapshot of the experiment, c) depth targeting and locking experiment result with six different jumping distance from 10 µm to 60 µm.

Fig. 6

Averages of RMSE with standard deviation error bar of the bovine retina experiment. The x-axis labels mean jumping distance and the y-axis indicates the means and standard deviations of the measurement specified in Table 1.