Surgical instrument segmentation in laparoscopic image sequences can be utilized for a variety of applications during surgical procedures. Recent studies have shown that deep learning-based methods produce competitive results in surgical instrument segmentation. Difficulties, however, lie in the limited number of training datasets involving surgical instruments in laparoscopic image frames. Even though there are publicly available pixelwise training datasets along with trained models from the Robotic Instrument Segmentation challenge, we are not able to relate them to laparoscopic image frames from different surgical scenarios without any pre- or postprocessing. This is because they contain different instrument shapes, image backgrounds, and specular reflections, which implies laborious manual segmentation for training dataset generation. In this work, we propose a novel framework for semi-automated training dataset generation for the purpose of robust segmentation using deep learning. To generate training datasets in various surgical scenarios faster and more accurately, we utilize the publicly available trained model from the Robotic Instrument Segmentation challenge and then use the Watershed Segmentation-based method. For robust segmentation, we use a two-step approach: first, we obtain a coarse segmentation obtained from a deep convolutional neural network architecture, and then we refine the segmentation result via the GrabCut algorithm. Through experiments using four different laparoscopic image sequences, we demonstrate the ability of our proposed framework to provide robust segmentation quality.
The purpose of this work was to develop a clinically viable laparoscopic augmented reality (AR) system employing stereoscopic (3-D) vision, laparoscopic ultrasound (LUS), and electromagnetic (EM) tracking to achieve image registration. We investigated clinically feasible solutions to mount the EM sensors on the 3-D laparoscope and the LUS probe. This led to a solution of integrating an externally attached EM sensor near the imaging tip of the LUS probe, only slightly increasing the overall diameter of the probe. Likewise, a solution for mounting an EM sensor on the handle of the 3-D laparoscope was proposed. The spatial image-to-video registration accuracy of the AR system was measured to be 2.59±0.58 mm and 2.43±0.48 mm for the left- and right-eye channels, respectively. The AR system contributed 58-ms latency to stereoscopic visualization. We further performed an animal experiment to demonstrate the use of the system as a visualization approach for laparoscopic procedures. In conclusion, we have developed an integrated, compact, and EM tracking-based stereoscopic AR visualization system, which has the potential for clinical use. The system has been demonstrated to achieve clinically acceptable accuracy and latency. This work is a critical step toward clinical translation of AR visualization for laparoscopic procedures.
Accurate calibration of laparoscopic cameras is essential for enabling many surgical visualization and navigation technologies such as the ultrasound-augmented visualization system that we have developed for laparoscopic surgery. In addition to accuracy and robustness, there is a practical need for a fast and easy camera calibration method that can be performed on demand in the operating room (OR). Conventional camera calibration methods are not suitable for the OR use because they are lengthy and tedious. They require acquisition of multiple images of a target pattern in its entirety to produce satisfactory result. In this work, we evaluated the performance of a single-image camera calibration tool (<i>rdCalib</i>; Percieve3D, Coimbra, Portugal) featuring automatic detection of corner points in the image, whether partial or complete, of a custom target pattern. Intrinsic camera parameters of a 5-mm and a 10-mm standard Stryker® laparoscopes obtained using rdCalib and the well-accepted OpenCV camera calibration method were compared. Target registration error (TRE) as a measure of camera calibration accuracy for our optical tracking-based AR system was also compared between the two calibration methods. Based on our experiments, the single-image camera calibration yields consistent and accurate results (mean TRE = 1.18 ± 0.35 mm for the 5-mm scope and mean TRE = 1.13 ± 0.32 mm for the 10-mm scope), which are comparable to the results obtained using the OpenCV method with 30 images. The new single-image camera calibration method is promising to be applied to our augmented reality visualization system for laparoscopic surgery.
Probe or needle artifact detection in 3D scans gives an approximate location for the tools inserted, and is thus crucial in assisting many image-guided procedures. Conventional needle localization algorithms often start with cropped images, where unwanted parts of raw scans are cropped either manually or by applying pre-defined masks. In cryoablation, however, the number of probes used, the placement and direction of probe insertion, and the portions of abdomen scanned differs significantly from case to case, and probes are often constantly being adjusted during the Probe Placement Phase. These features greatly reduce the practicality of approaches based on image cropping. In this work, we present a fully Automatic Probe Artifact Detection method, APAD, that works directly on uncropped raw MRI images, taken during the Probe Placement Phase in 3Tesla MRI-guided cryoablation. The key idea of our method is to first locate an initial 2D line strip within a slice of the MR image which approximates the position and direction of the 3D probes bundle, noting that cryoprobes or biopsy needles create a signal void (black) artifact in MRI with a bright cylindrical border. With the initial 2D line, standard approaches to detect line structures such as the 3D Hough Transform can be applied to quickly detect each probe’s axis. By comparing with manually labeled probes, the analysis of 5 patient treatment cases of kidney cryoablation with varying probe placements demonstrates that our algorithm combined with standard 3D line detection is an accurate and robust method to detect probe artifacts.