Estimating the Joint Angles of an Articulated Microrobotic Instrument Using Optical Coherence Tomography

Pose estimation of surgical tools is necessary for controlling and manipulating the tool in confined surgical regions for minimally invasive surgeries. Several studies have explored the application of robot assisted surgeries which require an imaging system to track the tool and navigate it with high accuracy and precision. Optical Coherence Tomog-raphy (OCT) is an emerging volumetric imaging modality in minimally invasive robot assisted surgery. We present a marker-based computer vision algorithm to estimate the wrist and finger joint angles of a neurosurgical gripper tool from an OCT volume. The tool's joint angles lie on two perpendicular planes. Markers of 1 mm diameter are placed on the tool and then the gripper is imaged under the OCT system. The raw volumetric images are pre-processed by downsampling and applying thresholds. The markers in the OCT volume are then detected using 3D template matching. False positive detections are algorithmically omitted by evaluating the relative distances between the markers. Finally, the positions of these detected markers are used to estimate the joint angles. Our approach yields an average error of 2.20 and a standard deviation of 3.4° for the wrist joint
$(\theta_{1})$
. For the finger joint
$(\theta_{2})$
, it yields an average error of 2° and a standard deviation of 1.5°• The estimations are provided within 0.5 seconds on a PC with a 16 core CPU.