Visual measurement of suture strain for robotic surgery

Abstract

Minimally invasive surgical procedures offer advantages of smaller incisions, decreased hospital length of stay, and rapid postoperative recovery to the patient. Surgical robots improve access and visualization intraoperatively and have expanded the indications for minimally invasive procedures. A limitation of the DaVinci surgical robot is a lack of sensory feedback to the operative surgeon. Experienced robotic surgeons use visual interpretation of tissue and suture deformation as a surrogate for tactile feedback. A difficulty encountered during robotic surgery is maintaining adequate suture tension while tying knots or following a running anastomotic suture. Displaying suture strain in real time has potential to decrease the learning curve and improve the performance and safety of robotic surgical procedures. Conventional strain measurement methods involve installation of complex sensors on the robotic instruments. This paper presents a noninvasive video processing-based method to determine strain in surgical sutures. The method accurately calculates strain in suture by processing video from the existing surgical camera, making implementation uncomplicated. The video analysis method was developed and validated using video of suture strain standards on a servohydraulic testing system. The video-based suture strain algorithm is shown capable of measuring suture strains of 0.2% with subpixel resolution and proven reliability under various conditions.

Figure 1

Marked suture held by the two grippers of a surgery robot.

Figure 2

Flow chart of image processing algorithm.

Figure 3

Color channel selection of dark suture (the red channel gives best contrast for suture image in the inverted image frame).

Figure 4

Suture line and marker detection process. (a) A binary image resulting from Sobel edge detection operators. (b) Hough transform on binary image for line detection. (c) Detected suture is displayed on the video image and markers are selected (red circles). (d) Line profile showing chosen markers (black arrows). (e) The pattern for the chosen marker (red line) is matched in the line profile (green line).

Figure 5

Marker tracking: quadratic curve fitting.

Figure 6

Strain methods.

Figure 7

Snapshot of the video processing for strain measurement display.

Figure 8

Loading test experimental setup

Figure 9

Video processing estimates of strain compared to the actual values as recorded by the materials testing system. (a) Trapezoidal reference waveform. (b) Strain measurement results.

Figure 10

Failure strain of most common surgical sutures: Our measurement system detects a minimum of 0.2% strain in suture which is well below the strain to failure (13%) (This figure was reprinted from [10], with permission from Elsevier).

Figure 11

Comparison of strain measurement for different contrast markers for 0.2% strain loading cycle.

Figure 12

Comparison of measurement results with one-point and two-point tracking.