Camera-Based Optical-Fiber Tactile Sensor for Intraoperative Grasping Force Measurement

Abstract

Introduction During minimally invasive surgery, surgeons insert specially-designed instruments through a small incision into the patient’s body. Despite all the advantages of this procedure, surgeons do not have the natural tactile force feedback in the surgery. Tactile feedback helps the surgeon to apply an appropriate force to avoid tissue damage [1]. Therefore, researchers have proposed different optical force sensors to attach at the tip of surgical instruments for measuring the tool-tissue interaction force [2]. To this end, a new sensing principle, i.e., variable bending radius (VBR) for light intensity modulation (LIM) in optical-fiber-based sensor was proposed and experimentally validated in our previous study [3]. This method allowed further miniaturization of MIS force sensors and increased its sensitivity. In the current study, a miniaturized VBR-based force sensor was fabricated using rapid prototyping techniques. The sensor was mounted on a custom-designed surgical grasper and an RGB camera was used to measure the transmitting light through the optical fiber. Finally, the calibrated sensor was studied for multiple grasping tasks on freshly excised bovine tissue. Material and Method As depicted in Fig. 1, the sensor has a flexible shell with a semi-circular indenter on the top of the optical fiber. Upon any contact between the flexible shell and tissue, the shell bends downward and the semi-circular indenter applies a force F on the optical fiber. Since optical fiber is anchored from both ends, it behaves as an Euler-Bernoulli beam under an external contact force. The sensor was 3D printed with an SLA printer with flexible and transparent resins. In order to acquire the output light intensity at the end of the optical fiber, an image-based intensity measurement system was developed. To this end, a Logitech C920 camera was used and a projection was 3D printed. The chamber was designed in such a way that the projection spot of laser in optical fiber would be visible in camera as depicted in Fig. 2(a). Noh et al. [4] have shown that average greyscale values, σ, of the image would be proportional to the total intensity of the laser spot and was defined as (1) where σi is the greyscale value of i-th pixel, and n = 640x480 was the total number of pixels in each frame. Fig. 2(b) depicts the variation of brightness of the laser spot as the sinusoidal force is applied on the sensor. Due to the bending of optical fiber, light intensity modulation happens as external force increases. In previous studies showed that the percentage of power decay in fiber is non-linearly dependent on the applied external force F. Also due to the viscoelastic properties of the 3D printed materials, the power decay is dependent on the rate of application of F and which would show its effects on the rate of change in output power. Therefore, a nonlinear rate-dependent learning-based support vector regression (SVR) calibration was trained with the data from a sinusoidal compression test (0-2N) with 0.5, 1.0, and 2.0Hz frequency. Fig. 2(c) depicts the training test setup. To further assess the performance of the proposed sensor in real surgical procedures, an ex-vivo test on freshly excised bovine muscle was performed (Fig. 3(a)). Also, to have a reference for the contact force between the jaws of the sensor and tissue, a FlexiForce sensor was calibrated and attached to the upper jaw of the sensorized grasper. Afterward, the tissue was manually grasped in the MIS grasper, while reference force was recorded through an Arduino MEGA2560 interface. In parallel, the force was estimated using the SVR formula. Results and Conclusions During the sinusoidal compression tests, the average greyscale value of the image taken by the camera decreased from 27- to 2-unit for 0 to 2N compressive force. SVR estimations showed an adjusted-R2 = 0.94 with a root-mean-square error of 0.09N between the testing machine measurements and SVR estimations. Also, SVR force estimations in the ex-vivo test showed fair agreement with the force readings from FlexiForce sensor (Fig. 3 (b)). In the ex-vivo validation test, the minimum detectable force of the sensor was 0.14N and the average error was 0.11±0.06N. As indicated in [1], force sensors for MIS applications must show a minimum detectable force of 0.2N with an average error of less than 0.2N. Therefore, authors conclude that the proposed and fabricated sensor with the implemented camera-based intensity measurement and the employed learning-based calibration technique have satisfactorily contributed to an applicable optical MIS force sensor. To expand the scope of this study, it is suggested that the ex-vivo tests be performed in an aqueous environment to study the effects of electrolytic interferences. Also, robustness of the proposed sensor against the electromagnetic interferences is to be studied.