Reducing residual forces in spinal fusion using a custom-built rod bending machine.

BACKGROUND AND OBJECTIVE: As part of spinal fusion surgery, shaping the rod implant to align with the anatomy is a tedious, error-prone, and time-consuming manual process. Inadequately contoured rod implants introduce stress on the screw-bone interface of the pedicle screws, potentially leading to screw loosening or even pull-out. METHODS: We propose the first fully automated solution to the rod bending problem by leveraging the advantages of augmented reality and robotics. Augmented reality not only enables the surgeons to intraoperatively digitize the screw positions but also provides a human-computer interface to the wirelessly integrated custom-built rod bending machine. Furthermore, we introduce custom-built test rigs to quantify per screw absolute tensile/compressive residual forces on the screw-bone interface. Besides residual forces, we have evaluated the required bending times and reducer engagements, and compared our method to the freehand gold standard. RESULTS: We achieved a significant reduction of the average absolute residual forces from for the freehand gold standard to (p=0.0015) using the bending machine. Moreover, our bending machine reduced the average time to instrumentation per screw from to . Reducer engagements per rod were significantly decreased from an average of 1.00±1.14 to 0.11±0.32 (p=0.0037). CONCLUSION: The combination of augmented reality and robotics has the potential to improve surgical outcomes while minimizing the dependency on individual surgeon skill and dexterity.

Segmenting/Pre-Processing Data from Bone Screw Thread-Stripping Tests.

Bone screws must be appropriately tightened to achieve optimal patient outcomes. If over-torqued, the threads formed in the bone may break, compromising the strength of the fixation; and, if under-torqued, the screw may loosen over time, compromising the stability. Previous work has proposed a model-based system to automatically determine the optimal insertion torque. This system consists of a reverse-modelling step to determine strength, and a forward modelling step to determine maximum torque. These have previously been tested in isolation, however future work must test the combined system. To do so, the data must be segmented and pre-processed. This was done based on specific features of the recorded data. The methodology was tested on 50 screw-insertion data sets across 5 different materials. With the parameters used, all data sets were correctly segmented. This will form a basis for the further processing of the data and validating the combined systemClinical relevance: The system for torque limit determination must be tested in its entirety to properly asses its performance. This paper discusses some of the steps required to pre-process the data to make this assessment. If successful, this system may improve patient outcomes in orthopaedic surgery.

Improved distinct bone segmentation in upper-body CT through multi-resolution networks.

PURPOSE: Automated distinct bone segmentation from CT scans is widely used in planning and navigation workflows. U-Net variants are known to provide excellent results in supervised semantic segmentation. However, in distinct bone segmentation from upper-body CTs a large field of view and a computationally taxing 3D architecture are required. This leads to low-resolution results lacking detail or localisation errors due to missing spatial context when using high-resolution inputs. METHODS: We propose to solve this problem by using end-to-end trainable segmentation networks that combine several 3D U-Nets working at different resolutions. Our approach, which extends and generalizes HookNet and MRN, captures spatial information at a lower resolution and skips the encoded information to the target network, which operates on smaller high-resolution inputs. We evaluated our proposed architecture against single-resolution networks and performed an ablation study on information concatenation and the number of context networks. RESULTS: Our proposed best network achieves a median DSC of 0.86 taken over all 125 segmented bone classes and reduces the confusion among similar-looking bones in different locations. These results outperform our previously published 3D U-Net baseline results on the task and distinct bone segmentation results reported by other groups. CONCLUSION: The presented multi-resolution 3D U-Nets address current shortcomings in bone segmentation from upper-body CT scans by allowing for capturing a larger field of view while avoiding the cubic growth of the input pixels and intermediate computations that quickly outgrow the computational capacities in 3D. The approach thus improves the accuracy and efficiency of distinct bone segmentation from upper-body CT.

Automatic patient positioning based on robot rotational workspace for extended reality.

PURPOSE: Understanding the properties and aspects of the robotic system is essential to a successful medical intervention, as different capabilities and limits characterize each. Robot positioning is a crucial step in the surgical setup that ensures proper reachability to the desired port locations and facilitates docking procedures. This very demanding task requires much experience to master, especially with multiple trocars, increasing the barrier of entry for surgeons in training. METHODS: Previously, we demonstrated an Augmented Reality-based system to visualize the rotational workspace of the robotic system and proved it helps the surgical staff to optimize patient positioning for single-port interventions. In this work, we implemented a new algorithm to allow for an automatic, real-time robotic arm positioning for multiple ports. RESULTS: Our system, based on the rotational workspace data of the robotic arm and the set of trocar locations, can calculate the optimal position of the robotic arm in milliseconds for the positional and in seconds for the rotational workspace in virtual and augmented reality setups. CONCLUSIONS: Following the previous work, we extended our system to support multiple ports to cover a broader range of surgical procedures and introduced the automatic positioning component. Our solution can decrease the surgical setup time and eliminate the need to repositioning the robot mid-procedure and is suitable both for the preoperative planning step using VR and in the operating room-running on an AR headset.

Improved distinct bone segmentation from upper-body CT using binary-prediction-enhanced multi-class inference.

PURPOSE: Automated distinct bone segmentation has many applications in planning and navigation tasks. 3D U-Nets have previously been used to segment distinct bones in the upper body, but their performance is not yet optimal. Their most substantial source of error lies not in confusing one bone for another, but in confusing background with bone-tissue. METHODS: In this work, we propose binary-prediction-enhanced multi-class (BEM) inference, which takes into account an additional binary background/bone-tissue prediction, to improve the multi-class distinct bone segmentation. We evaluate the method using different ways of obtaining the binary prediction, contrasting a two-stage approach to four networks with two segmentation heads. We perform our experiments on two datasets: An in-house dataset comprising 16 upper-body CT scans with voxelwise labelling into 126 distinct classes, and a public dataset containing 50 synthetic CT scans, with 41 different classes. RESULTS: The most successful network with two segmentation heads achieves a class-median Dice coefficient of 0.85 on cross-validation with the upper-body CT dataset. These results outperform both our previously published 3D U-Net baseline with standard inference, and previously reported results from other groups. On the synthetic dataset, we also obtain improved results when using BEM-inference. CONCLUSION: Using a binary bone-tissue/background prediction as guidance during inference improves distinct bone segmentation from upper-body CT scans and from the synthetic dataset. The results are robust to multiple ways of obtaining the bone-tissue segmentation and hold for the two-stage approach as well as for networks with two segmentation heads.

Fabrication and Characterization of a Flexible FBG-Based Shape Sensor Using Single-Mode Fibers.

Minimally invasive surgical procedures have become the preferable option, as the recovery period and the risk of infections are significantly lower than traditional surgeries. However, the main challenge in using flexible tools for minimal surgical interventions is the lack of precise feedback on their shape and tip position inside the patient's body. Shape sensors based on fiber Bragg gratings (FBGs) can provide accurate shape information depending on their design. One of the most common configurations in FBG-based shape sensors is to attach three single-mode optical fibers with arrays of FBGs in a triangular fashion around a substrate. Usually, the selected substrates dominate the bending stiffness of the sensor probe, as they have a larger diameter and show less flexibility compared to the optical fibers. Although sensors with this configuration can accurately estimate the shape, they cannot be implemented in flexible endoscopes where large deflections are expected. This paper investigates the shape sensor's performance when using a superelastic substrate with a small diameter instead of a substrate with dominating bending stiffness. A generalized model is also designed for characterizing this type of flexible FBG-based shape sensor. Moreover, we evaluated the sensor in single and multi-bend deformations using two shape reconstruction methods.

Quantitative Evaluation of the Thickness of the Available Manipulation Volume Inside the Knee Joint Capsule for Minimally Invasive Robotic Unicondylar Knee Arthroplasty.

OBJECTIVE: Developing robotic tools that introduce substantial changes in the surgical workflow is challenging because quantitative requirements are missing. Experiments on cadavers can provide valuable information to derive workspace requirements, tool size, and surgical workflow. This work aimed to quantify the volume inside the knee joint available for manipulation of minimally invasive robotic surgical tools. In particular, we aim to develop a novel procedure for minimally invasive unicompartmental knee arthroplasty (UKA) using a robotic laser-cutting tool. METHODS: Contrast solution was injected into nine cadaveric knees and computed tomography scans were performed to evaluate the tool manipulation volume inside the knee joints. The volume and distribution of the contrast solution inside the knee joints were analyzed with respect to the femur, tibia, and the anatomical locations that need to be reached by a laser-cutting tool to perform bone resection for a standard UKA implant. RESULTS: Quantitative information was determined about the tool manipulation volume inside these nine knee joints and its distribution around the cutting lines required for a standard implant. CONCLUSION: Based on the volume distribution, we could suggest a possible workflow for minimally invasive UKA, which provides a large manipulation volume, and deducted that for the proposed workflow, an instrument with a thickness of 5-8 mm should be feasible. SIGNIFICANCE: We present quantitative information on the three-dimensional distribution of the maximally available volume inside the knee joint. Such quantitative information lays the basis for developing surgical tools that introduce substantial changes in the surgical workflow.

The Functional Rotational Workspace of a Human-Robot System can be Influenced by Adjusting the Telemanipulator Handle Orientation.

The handle design of telemanipulation master devices has not been extensively studied so far. However, the master device handle is an integral part of the robotic system through which the user interacts with the system. Previous work showed that the size and shape of the functional rotational workspace of the human-robot system and its usability are influenced by the design of the master device handle. Still, in certain situations, e.g., due to user preference, a specific grasp type handle might be desired. Therefore, in this article, we provide a systematic approach on how to assess and adjust the functional rotational workspace of a human-robot system. We investigated the functional rotational workspace with two exemplary grasp type handles and two different mounting orientations for each handle. The results showed that by adapting the handle orientation in the home configuration of the telemanipulator, the functional rotational workspace of the human-robot system can be adjusted systematically to cover more of the mechanical workspace of the master device. Finally, we deduct recommendations on how to choose and adjust a telemanipulator handle.

Force quantification and simulation of pedicle screw tract palpation using direct visuo-haptic volume rendering.

PURPOSE: We present a feasibility study for the visuo-haptic simulation of pedicle screw tract palpation in virtual reality, using an approach that requires no manual processing or segmentation of the volumetric medical data set. METHODS: In a first experiment, we quantified the forces and torques present during the palpation of a pedicle screw tract in a real boar vertebra. We equipped a ball-tipped pedicle probe with a 6-axis force/torque sensor and a motion capture marker cluster. We simultaneously recorded the pose of the probe relative to the vertebra and measured the generated forces and torques during palpation. This allowed us replaying the recorded palpation movements in our simulator and to fine-tune the haptic rendering to approximate the measured forces and torques. In a second experiment, we asked two neurosurgeons to palpate a virtual version of the same vertebra in our simulator, while we logged the forces and torques sent to the haptic device. RESULTS: In the experiments with the real vertebra, the maximum measured force along the longitudinal axis of the probe was 7.78 N and the maximum measured bending torque was 0.13 Nm. In an offline simulation of the motion of the pedicle probe recorded during the palpation of a real pedicle screw tract, our approach generated forces and torques that were similar in magnitude and progression to the measured ones. When surgeons tested our simulator, the distributions of the computed forces and torques were similar to the measured ones; however, higher forces and torques occurred more frequently. CONCLUSIONS: We demonstrated the suitability of direct visual and haptic volume rendering to simulate a specific surgical procedure. Our approach of fine-tuning the simulation by measuring the forces and torques that are prevalent while palpating a real vertebra produced promising results.

Optoacoustic Tissue Differentiation Using a Mach-Zehnder Interferometer.

Laser osteotomy offers a way to make precise and less traumatic cuts smaller than conventional mechanical bone surgery tools. To fully exploit the advantages of laser osteotomy over conventional osteotomy, real-time feedback to differentiate the hard bone from the surrounding soft tissues is desired. In this study, we differentiated various tissue types-hard and soft bone, fat, muscle, and skin tissues from five proximal and distal fresh porcine femurs-based on cutting sounds. For laser ablation, an Nd:YAG laser was used to create ten craters on the surface of each proximal and distal femurs. For sound recording, the probing beam of a Mach-Zehnder interferometer was placed 5 cm away from each ablation site. For offline tissue differentiation, we investigated the measurements by looking at the amplitude frequency band between 0.83 and 1.25 MHz, which provides the least average classification error. Then, we used principal component analysis to reduce the dimensionality and the 95% confidence ellipsoid (Mahalanobis distance) method to differentiate between tissues based on the acoustic shock wave. A set of 14 400 data points, measured from ten craters in four proximal and distal femurs, was used as "training data," while a set of 3600 data points from ten craters in the remaining proximal and distal femurs was considered as "testing data." As is seen in the confusion matrix, the experimental-based scores of hard and soft bones, fat, muscles, and skin yielded average classification errors (with leave-one-out cross validation) of 0.11%, 57.69%, 0.06%, 0.14%, and 2.92%, respectively. The results of this study demonstrate a promising technique for differentiating tissues during laser osteotomy.

Laser-induced breakdown spectroscopy as a potential tool for autocarbonization detection in laserosteotomy.

In laserosteotomy, it is vital to avoid thermal damage of the surrounding tissue, such as carbonization, since carbonization does not only deteriorate the ablation efficiency but also prolongs the healing process. The state-of-the-art method to avoid carbonization is irrigation systems; however, it is difficult to determine the desired flow rate of the air and cooling water based on previous experiments without online monitoring of the bone surface. Lack of such feedback during the ablation process can cause carbonization in case of a possible error in the irrigation system or slow down the cutting process when irrigating with too much cooling water. The aim of this paper is to examine laser-induced breakdown spectroscopy as a potential tool for autocarbonization detection in laserosteotomy. By monitoring the laser-driven plasma generated during nanosecond pulse ablation of porcine bone samples, carbonization is hypothesized to be detectable. For this, the collected spectra were analyzed based on variation of a specific pair of emission line ratios in both groups of samples: normal and carbonized bone. The results confirmed a high accuracy of over 95% in classifying normal and carbonized bone.