Volume 8, Issue 1 - Proceedings of the CURAC 2022 Annual Meeting, Karlsruhe, September 15–17, 2022

Minimal-invasive procedures come with significant advantages for the patient. They also come with problems as the navigation/guidance of the devices to a target location is either based on pre-operatively acquired images and then performed free-hand or is accompanied by intraoperative imaging such as MRI or CT that is expensive, complicated and produces artifacts. Using robotic systems for moving and guiding these interventional and therapeutic devices adds additional issues like lack of palpation sensation and missing tissue feedback. While it is possible to add sensors to the distal tip, this creates other obstacles concerning reduced functionality, cables, sterility issues and added complexity and cost. We propose to use a proximally attached audio sensor to record the tissue tool interaction and provide real-time feedback to the clinician. This paper reports on initial attempts to use this technology with robotic arms for surface characterization and interventional vascular procedures that gain increased attention in combination with robotic devices. In summary, Proximal Audio Sensing could be a versatile, cost-effective and powerful tool to guide minimally invasive needle interventions and enable (semi-) autonomous robot-assisted surgery.

Laparoscopic cholecystectomy (LCHE) is a widely employed model for surgical instrument and phase recognition in the field of machine learning (ML), with the latter being assigned to identify critical events and to avoid complications. Although ML algorithms have been proven to be effective for this instance and in selected patients, it is questionable whether patients receiving LCHE in daily clinical routine would actually benefit from adverse event prediction by ML applications. We believe, that the statistical problem of low prevalence (PREV) of potential adverse events in an unselected population and consequential low diagnostic yield was not considered adequately in recent research. Therefore, we performed a query to the G-DRG (German Diagnosis Related Groups) database of the German Federal Statistical Office with the aim to calculate prevalence of surgical and postoperative adverse events coming along with LCHE. The results enable an estimation of positive (PPV) and negative (NPV) predictive values hypothetically achievable by ML applications aiming to predict an adverse surgical course.

Training clipping of intracranial aneurysms is challenging due to less frequent interventions because of minimalinvasive methods, ethical aspects regarding cadaver training and consumption of resources when using 3D printed models. By proposing an immersive virtual reality training system, we make use of increased motivation, engagement and realism when using a virtual operating room. In this simulation, a selected microsurgical clip can be applied at the aneurysm neck. Before closing the clip and deforming the vessels, the affected area is visualized to assess the clip position. Our qualitative evaluation with two neurosurgeons with different levels of experience indicates benefits such as increased motivation, presence, and the possibility to try out different strategies. However, some surgical steps can be refined to increase realism and learning effect, and interactions can be further improved. The proposed training system benefits from training by trial and error in an engaging environment leading to an improved training experience.

In contrast to unstructured meshes, structured meshes yield faster simulation results for bio-medical simulations, but are very time-consuming to create. A preprocessing step in the generation of structured meshes is the manual construction of a blockstructure approximating the vessel. Here, we present an automatic centerline calculation and blockstructure generation to reduce the user effort and time of structured mesh generation. The centerline is detected as points in between opposite faces. Based on the centerline, cross sections are determined and a blockstructure which approximates the vessel is automatically generated. The centerline detection does not require time-consuming user input and meshes with more than 195,000 vertices are processed in less than 160 seconds. The results of the presented automatic centerline detection are compared to a centerline with manual input generated by the widely used vmtk tool. The centerlines are similar, small differences occur at bifurcation and at the aneurysm.

The task of medical image classification is increasingly supported by algorithms. Deep learning methods like convolutional neural networks (CNNs) show superior performance in medical image analysis but need a high-quality training dataset with a large number of annotated samples. Particularly in the medical domain, the availability of such datasets is rare due to data privacy or the lack of data sharing practices among institutes. Generative adversarial networks (GANs) are able to generate high quality synthetic images. This work investigates the capabilities of different state-of-the-art GAN architectures in generating realistic breast ultrasound images if only a small amount of training data is available. In a second step, these synthetic images are used to augment the real ultrasound image dataset utilized for training CNNs. The training of both GANs and CNNs is conducted with systematically reduced dataset sizes. The GAN architectures are capable of generating realistic ultrasound images. GANs using data augmentation techniques outperform the baseline Style- GAN2 with respect to the Frechet Inception distance by up to 64.2%. CNN models trained with additional synthetic data outperform the baseline CNN model using only real data for training by up to 15.3% with respect to the F1 score, especially for datasets containing less than 100 images. As a conclusion, GANs can successfully be used to generate synthetic ultrasound images of high quality and diversity, improve classification performance of CNNs and thus provide a benefit to computer-aided diagnostics.

Cardiovascular diseases are the main cause of death worldwide. State-of-the-art treatment often includes the process of navigating endovascular instruments through the vasculature. Automation of the procedure receives much attention lately and may increase treatment quality and unburden clinicians. However, in order to ensure the patient’s safety the endovascular device needs to be steered carefully through the body. In this work, we present a collection of medical criteria that are considered by physicians during an intervention and that can be evaluated automatically enabling a highly objective assessment. Additionally, we trained an autonomous controller with deep reinforcement learning to gently navigate within a 2D simulation of an aortic arch. Among others, the controller reduced the maximum and mean contact force applied to the vessel walls by 43% and 29%, respectively.

This work presents a sterile concept for 3D cameras within sterile surgical environments. In the digital operating room (OR), such cameras can serve as a valuable data source for cognitive workflow assistance systems, e.g decision or mechatronic support systems. One recent example are robotic assistants for instrument handling, such as the robotic scrub nurse currently developed in the framework of the SASHA-OR research project1. In this context, we detect laparoscopic instruments and the surgical environment with a 3D camera, whereby hygienic requirements need to be met. Using a Zivid Two sensor, we generated point clouds of the laparoscopic instruments located in an instrument holder and a drop zone. We compared the effect of using different pane types and thicknesses for the sterile camera enclosure and compared the performance with and without protective pane in terms of the point cloud accuracy. When analyzing multiple pane types, polymethyl methacrylate with 0.5 mm thickness (PMMA 0.5) provided the best results. At a scan distance of 560 mm to the surface center, which is required for the complete acquisition of a laparoscopic instrument, PMMA 0.5 achieved the smallest Chamfer distance (CD) values for both the scans with the laparoscopic instruments in the instrument holder (0.23 ± 1.52 mm) and in the drop zone (0.12 ± 0.25 mm).

Machine learning-based solutions rely heavily on the quality and quantity of the training data. In the medical domain, the main challenge is to acquire rich and diverse annotated datasets for training. We propose to decrease the annotation efforts and further diversify the dataset by introducing an annotation-efficient learning workflow. Instead of costly pixel-level annotation, we require only image-level labels as the remainder is covered by simulation. Thus, we obtain a large-scale dataset with realistic images and accurate ground truth annotations. We use this dataset for the instrument localization activity task together with a studentteacher approach. We demonstrate the benefits of our workflow compared to state-of-the-art methods in instrument localization that are trained only on clinical datasets, which are fully annotated by human experts.

Radiographs are a versatile diagnostic tool for the detection and assessment of pathologies, for treatment planning or for navigation and localization purposes in clinical interventions. However, their interpretation and assessment by radiologists can be tedious and error-prone. Thus, a wide variety of deep learning methods have been proposed to support radiologists interpreting radiographs. Mostly, these approaches rely on convolutional neural networks (CNN) to extract features from images. Especially for the multi-label classification of pathologies on chest radiographs (Chest X-Rays, CXR), CNNs have proven to be well suited. On the Contrary, Vision Transformers (ViTs) have not been applied to this task despite their high classification performance on generic images and interpretable local saliency maps which could add value to clinical interventions. ViTs do not rely on convolutions but on patch-based self-attention and in contrast to CNNs, no prior knowledge of local connectivity is present. While this leads to increased capacity, ViTs typically require an excessive amount of training data which represents a hurdle in the medical domain as high costs are associated with collecting large medical data sets. In this work, we systematically compare the classification performance of ViTs and CNNs for different data set sizes and evaluate more data-efficient ViT variants (DeiT). Our results show that while the performance between ViTs and CNNs is on par with a small benefit for ViTs, DeiTs outperform the former if a reasonably large data set is available for training.

Purpose The demand for telemedicine is continuously increasing. However, many medical examination techniques such as auscultation are based on manual execution and subjective assessment by each individual doctor. For the increasing trend of telemedicine, we have developed an electric percussion device that examines in a standardized way. Methods This study evaluated the new approach of using motorized percussion against the gold standard of manual percussion. An examination was carried out on a healthy volunteer using both techniques. Thereby, occurring sounds from both pulmonary and non-pulmonary areas were recorded as the primary goal of percussion is to assess the size of the lung. Recordings were cut into individual samples and randomized within each examination variant. Results When discriminating between ventral and dorsal, and manual and motorized percussion, the outcome of the 28-participant anonymous survey showed that the noisesuppressed ventral motorized percussion with 93.2% (10.2) correctly identified percussion points surpassed all other conducted percussion techniques. Conclusion Making a distinction between pulmonary volume and exterior areas is very reliable with the motorized percussion. In order to be able to describe the clinical picture precisely in the future, further evaluations are necessary.

Tissue elasticity contains important information for physicians in diagnosis and treatment, and, e.g., can help in tumor detection because tumors are stiffer than healthy tissue. Ultrasound shear wave elastography imaging (US-SWEI) can be used to estimate tissue stiffness by measuring the velocity of induced shear waves. Commonly, a linear US probe is used to track shear waves at a high imaging frequency in 2D. Realtime US-SWEI is limited by the required time for data processing. Hence, reducing the imaging field of view (FOV) is beneficial as it decreases the size of the acquired data and thereby the acquisition, transfer and processing time. However, a decrease in the FOV has the disadvantage that shear waves are tracked over a smaller distance and thus, also fewer sampling points are available for velocity estimation. This trade-off between a smaller FOV and thus, a smaller data size, and the accuracy of shear wave velocity estimation is investigated in this work. For this purpose, shear waves were tracked with a linear US probe in gelatin phantoms with four different stiffness values. During data processing, we reduced the FOV virtually from 38.1mm to 2.1mm. It was found that a reduction of the FOV to 4.5mm leads to an overestimation of up to five times larger shear wave velocities but still allows to distinguish between phantoms of different stiffness. However, not all estimated velocity values could be clearly assigned to the correct stiffness value. The smallest studied FOV of 2.1mm was not sufficient for distinguishing between the phantoms anymore.

In laparoscopic liver surgery (LLS) image-guided navigation systems could support the surgeon by providing subsurface information such as the positions of tumors and vessels. For this purpose, one option is to perform a registration of preoperative 3D data and 3D surface patches reconstructed from laparoscopic images. Part of an automatic 3D registration pipeline is the feature description, which takes into account various geometric and spatial information. Since there is no leading feature descriptor in the field of LLS, two feature descriptors are compared in this paper: The Fast Point Feature Histogram (FPFH) and Triple Orthogonal Local Depth Images (TOLDI). To evaluate their performance, three perturbations were induced: varying surface patch sizes, spatial displacement, and Gaussian deformation. Registration was performed using the RANSAC algorithm. FPFH outperformed TOLDI for small surface patches and in case of Gaussian deformations in terms of registration accuracy. In contrast, TOLDI showed lower registration errors for patches with spatial displacement. While developing a 3D-3D registration pipeline, the choice of the feature descriptor is of importance, consequently a careful choice suitable for the application in LLS is necessary.

Robot-assisted surgery (RAS) has been widely accepted in the clinical routine around the world with a focus on urology, gynecology and general surgery. Even transoral robotic surgery (TORS) is applied on a regular basis in some specialized medical centers. However, the anatomical constraints and the delicate procedures at the anterior skull base prevent the introduction of RAS until today. We have performed an extended literature research on robot-assisted skull base surgery and added our own pre-clinical experiences to the list of publications. The overview includes available systems, pre-clinical and clinical applications, specific limitations, remaining challenges and an outlook into future developments.

Minimally invasive surgery is the standard for many abdominal interventions, with an increasing use of telemanipulated robots. As collaborative robots enter the field of medical interventions, their intuitive control needs to be addressed. Augmented reality can thereby support a surgeon by representing the surgical scene in a natural way. In this work, an augmented reality based robot control for laparoscopic cholecystectomy is presented. A user can interact with the virtual scene to clip the cystic duct and artery as well as to manipulate the deformable gallbladder. An evaluation was performed based on the SurgTLX and system usability scale.

Acquiring reproducible ultrasound images of high quality is challenging in ultrasound imaging. While physicians can rely on their experience, robot-assisted systems must be able to automatically align the ultrasound probe with the correct orientation. This paper describes a method to align the central axis of the probe to the surface normal at the point of contact. This is done by analyzing the area of ultrasound volumes directly below the transducer of the probe. A convolutional neural network is trained to estimate the inclination of the probe orientation towards the direction of the surface normal. Experiments on two different phantoms indicate that the mean absolute angle error between the estimated rotation and the ground truth surface normal are 5.0 ± 2.8∘. The method is able to keep the probe-surface contact continuous and the results indicate that the method is invariant to anatomical structures.

Medical interventions are often guided by optical tracking systems and optical coherence tomography has shown promising results for markerless tracking of soft tissue. The high spatial resolution and subsurface information contain valuable information about the underlying tissue structure and tracking of certain target structures is in principle possible. However, the small field-of-view complicates the selection of suitable regions-of-interest for tracking. Therefore, we extend an experimental setup and perform volumetric surface scanning of target structures to enlarge the field-of-view. We show that the setup allows for data acquisition and that precise merging of the volumes is possible with mean absolute errors from 0.041mm to 0.097mm.

The manipulation of flexible endoscopes is a procedure that requires great dexterity since it requires the synchronization and use of both hands in parallel. Imprecise handling during gastroendoscopy could harm the digestive tract. Our solution allows the physician to use only one hand to wirelessly control the forward, backward, and tip bending motion. The proposed system provides endoscopic vision and tactile impact force sensing at the tip to detect the force applied to tissue and thus avoid damage. We experimentally evaluate the handling of the robotic system in open space and inside a medical phantom. The results revealed a training effect with less time demand for task completion and reduction of average impact force after only 5 runs. The proposed system was successfully controlled using one hand and, together with the force information, could enhance the physician’s experience during endoscopy. Future work will address axial control and an intensive user study with clinical experts.

Touchless interaction is popular in the medical domain because it maintains sterility and ensures physicians’ autonomy. Evaluating these technologies, however, proves difficult due to technical and human hurdles. Virtual reality leaves these limitations behind and allows for the exploration of promising concepts by simulating an environment and the interactions that takes place within it.We present a virtual radiology suite in the context of needle-based MR-interventions to evaluate touchless interactions. Hand and foot inputs were implemented on a custom interface and evaluated in a user study (n= 16). Results show that activating the system and manipulating values was faster with foot input. However, multimodal interaction is preferable because it is less demanding.

The rapid development of available hard- and software for computer-assisted or augmented reality (AR) guided interventions creates a need for fast and inexpensive prototyping environments. However, intraoperative tracking systems in particular represent a high cost threshold. Therefore, this work presents a low-cost tracking method based on a conventional RGB camera. Here, a combined approach of multiple image targets and 3D object target recognition is implemented. The system is evaluated with a systematic accuracy assessment analyzing a total of 385 3D positions. On average, a deviation of 15,69+-9,95 mm was measured. In addition, a prototypical AR-based needle navigation visualization was developed using Microsoft HoloLens 2. This system’s feasibility and usability was evaluated positively in a pilot study (n=3).

The paper describes how eye-tracking can be used to explore electronic patient records (EPR) in a sterile environment. As an information display, we used a system that we developed for the presentation of patient data and for supporting surgical hand disinfection. The eye-tracking was performed using the Tobii Eye Tracker 4C, and the connection between the eye-tracker and the HTML website was realized using the Tobii EyeX Chrome Extension. Interactions with the EPR are triggered by fixations of icons. The interaction was working as intended, but test persons reported a high mental load while using the system.

Ultrasound is a widely used imaging technique and is appreciated for its non-invasiveness, absence of radiation, widespread availability, and compact equipment. Ergonomic difficulties in manual handling of the probe could be enhanced by a robotic controlled ultrasound. The paper addresses the development of such a system which enables remote operation of a ultrasound probe and includes haptic force feedback as well as video conferencing components for visual feedback. The development process followed a user-centered approach by investigating needs of potential end-users. Preliminary results demonstrated the functionality of the developed system for generating medical image data under laboratory conditions.

In this publication, methods are presented to improve predictions of perioral soft-tissue changes following the expansion of the dental arches during orthodontic therapy. Acrylic veneers with different thicknesses were reversibly attached to the buccal surfaces of the upper and lower incisors to simulate their protrusion. The resulting morphological changes of the perioral soft-tissue surface were determined by 3D face scans. Experimentally-determined 3D soft-tissue changes are compared to numerical predictions using detailed finite-element (FE) models of the face of two individuals differing in the body mass index (BMI). The results suggest that common estimates of material constants used by the detailed and individualized FE models might be sufficient to explain absolute soft-tissue displacements although differences occurred between experimental and modeling results regarding the directions of displacements. The aim of this investigation is to create predictions of post-treatment appearance that are helpful for therapy planning.

Developing autonomous endovascular robotic systems requires physical testbeds to test control algorithms. Typically, such testbeds comprise of several hard- and software components along with a way of having these components communicate with each other. Building such a testbed is a multidisciplinary task which can be beyond the scope of expertise for research groups. The goal of this work is to facilitate setting up such testbeds in two ways: First, we propose a testbed architecture that allows to develop tracking, control and instrument manipulation systems separately by utilizing the ROS2 communication protocol. Secondly, we present a reliable yet straightforward to implement tracking algorithm for endovascular instruments that is built using only open-source software packages. The tracking algorithm is evaluated using both video camera and x-ray imaging and is found to meet the requirements for real time control algorithms. Furthermore, we show an example of the proposed modular testbed architecture as it is used in our lab. Both the modular testbed architecture and the open-source tracking algorithm may serve as helpful building blocks for other researchers in the field seeking to evaluate their control algorithms on physical testbeds.

Medical drilling processes, especially in the area of the otobase, pose a high risk to the patient due to the heat generated and introduced into the tissue. Two measures of reducing the heat are the parameters of the drilling process and the optimisation of the drill geometry. A spiral drill and a singlelip drill are evaluated for their suitability as medical drills in skull base surgery with the objective of integrating a temperature sensor in the tip of the drill. The two drills were chosen because of their promising heat generation properties in pretesting. In this paper the chip formation behaviour and drilling channel quality are analysed. The single-lip drill shows lack in precision of the drilling channel and unsuitable bone chip transport, as material powder is left in the drilling channel. In contrast, the spiral drill shows a good transport of the chips and the drilling channel follows the planned path. As a result, the spiral drill is chosen for the further development of a drill with integrated temperature sensor for minimally invasive drilling.

A growing body of evidence supports Virtual Reality (VR) as an effective and safe strategy for management of pain and stress associated with medical procedures in both adults and children. We therefore initiated a feasibility study to investigate the effect of VR on pain, stress, and anxiety during elective surgery, e.g. implantation of a central-venous port catheter, hypothesizing that VR can reduce intraoperative pain, stress and anxiety of the patient. In this manuscript, the preliminary results of the first 20 (out of 6o planned) patients are presented. Baseline pain characteristics did not differ between the two study groups (VR group (n=10) and standard (no VR device) group (n=10)). System usability (“easy to use”, “easy to learn” and “safe”) was rated “good - very good” by the study participants. Selfassessment of anxiety components (Y-6 item questionnaire) revealed a calming (3.3 ± 0.5 vs. 2.4 ± 0.5, P= 0.009) and relaxing (2.7 ± 1.2 vs. 1.8 ± 0.4; P=0.09) effect of the VR device. Evaluation of pain level (Short form McGill questionnaire) during the procedure revealed a lower pain intensity (VAS) level (17.5 ± 12.1 vs. 19.5 ± 10.6; P= 0.834) and present pain intensity (PPI) score (0.9 ± 0.6 vs. 1.0 ± 0.5; P= 0.841) in the VR group Preliminary data of our feasibility study indicates a positive effect of VR towards reduction of pain and stress in patients undergoing minor surgery in local anaesthesia. However, further data is needed to substantiate these results.

Accurate measurement of interaction forces during robot-assisted surgery requires compact force sensing modalities in the surgical tools, thus might add considerable cost to the setup. Measuring the motor current to estimate gripping forces, is an advantageous approach since no expensive force sensor is needed. In this paper, a mechanical interface is presented, which allows actuating conventional articulated instruments for robot-assisted surgery. The interface features the estimation of static gripping forces at the instrument’s tip based on the motor current. The evaluation shows reproducible results, and the current-based approach seems to be a cost-efficient way to estimate gripping forces.

With the progress of technology in modern hospitals, an intelligent perioperative situation recognition will gain more relevance due to its potential to substantially improve surgical workflows by providing situation knowledge in real-time. Such knowledge can be extracted from image data by machine learning techniques but poses a privacy threat to the staff’s and patients’ personal data. De-identification is a possible solution for removing visual sensitive information. In this work, we developed a YOLO v3 based prototype to detect sensitive areas in the image in real-time. These are then deidentified using common image obfuscation techniques. Our approach shows that it is principle suitable for de-identifying sensitive data in OR images and contributes to a privacyrespectful way of processing in the context of situation recognition in the OR.

When documenting colonoscopic examinations, the challenge arises, that the exact location of the detected lesions (such as polyps) can usually only be determined with low precision. In order to improve the image-based documentation of such lesions with respect to their location and context, a system is currently investigated and developed, allowing the creation of a panoramic map of the interior of the colon’s lumen during an endoscopic examination. To evaluate the proposed algorithms, a 3D-printed colon phantom with known geometry has been developed. Furthermore, to obtain image sequences with precise and known motion during a simulated examination for test purposes, a robot was used to perform the insertion and withdrawal of the endoscope with constant speed and a steady viewing angle. The thus obtained image data is used to generate an ideal panoramic map using our the proposed stitching algorithm. Hence, it is possible to obtain reference maps of the phantom, which can be used to evaluate the quality of the algorithm.

Navigation systems play an increasingly important role in minimally invasive surgery (MIS) by mitigating the problems rising from the decoupling of hand-eye movement of the surgeon. Many of these systems suffer from a high dependency on external optical tracking systems that require a constant line-of-sight to the optical markers being tracked. Simultaneous localization and mapping (SLAM) algorithms allow tracking the endoscope in cases where optical tracking fails due to the cluttered environment in the operating room. To ensure a correct camera pose estimate and to correct for drift, a recognition of previously visited locations (loop closures) is essential. We propose a method for location recognition in a minimally invasive scenario that only requires a stereo endoscope and an inertial measurement unit (IMU).We use a hierarchical bag-of-visual-words (BoW) algorithm that saves compact image representations and enables querying for matching images. A two-staged consistency check using a random sample consensus (RANSAC) and the data measured by the IMU ensure a high matching precision.

An increasing number of colonoscopic guidance and assistance systems rely on machine learning algorithms which require a large amount of high-quality training data. In order to ensure high performance, the latter has to resemble a substantial portion of possible configurations. This particularly addresses varying anatomy, mucosa appearance and image sensor characteristics which are likely deteriorated by motion blur and inadequate illumination. The limited amount of readily available training data hampers to account for all of these possible configurations which results in reduced generalization capabilities of machine learning models. We propose an exemplary solution for synthesizing colonoscopy videos with substantial appearance and anatomical variations which enables to learn discriminative domain-randomized representations of the interior colon while mimicking real-world settings.

For hemodynamic simulations of intracranial aneurysms boundary conditions (BC) are required. In most cases, these are not patient-specific and thus do not reflect the real flow conditions in the patient. This study investigates the influence of patient-specific inflow BC on intra-aneurysmal hemodynamics. The focus lies on gender and age variations of the patients. To asses the impact, four different inflow curves representing the velocity profile of the inflow over one cardiac cycle is modeled. These four inflow BC are varied in the simulations of each aneurysm from selected subgroups. From the results of the simulations, the hemodynamic parameters are determined for each inflow BC and the percent differences between inflow BC are determined. The results show that the hemodynamic parameters are not robust to varying inflow BC. It can be seen that age has more influence on the hemodynamic parameters than gender. This study demonstrates the dependence of valid hemodynamic parameters on realistic inflow BC. Thus, if available, patient-specific inflow curves are recommended.

7T TOF MRI scans provide high resolution images of intracranial vasculature. When segmented, the Circle of Willis is detailed and thus opens up new possibilities, in research but also in education. We propose a segmentation pipeline for the Circle ofWillis, and introduce a prototype that enables exploration of not just the entire Circle of Willis, but also of its centerline, in an immersive VR enviroment. In our prototype, the model can be freely rotated, placed and scaled. A qualitative evaluation was performed with two experienced neuroradiologists, who rated the prototype and its potential positively.

Intraoperative imaging can assist neurosurgeons to define brain tumours and other surrounding brain structures. Interventional ultrasound (iUS) is a convenient modality with fast scan times. However, iUS data may suffer from noise and artefacts which limit their interpretation during brain surgery. In this work, we use two deep learning networks, namely UNet and TransUNet, to make automatic and accurate segmentation of the brain tumour in iUS data. Experiments were conducted on a dataset of 27 iUS volumes. The outcomes show that using a transformer with UNet is advantageous providing an efficient segmentation modelling long-range dependencies between each iUS image. In particular, the enhanced TransUNet was able to predict cavity segmentation in iUS data with an inference rate of more than 125 FPS. These promising results suggest that deep learning networks can be successfully deployed to assist neurosurgeons in the operating room.

Varying device stiffness on purpose can provide safety and performance improvements in interventions involving flexible minimal-invasive surgical robotic devices. For example, when flexible robotic devices are used for inserting minimal-invasive tools along curved paths into the knee joint for arthroscopy, a high device stiffness can allow precise path following where desirable. In contrast, when the minimalinvasive flexible device has reached its desired location inside the patient’s body, decoupling the surgical tool (end-effector) from the flexible robotic insertion device can be beneficial in avoiding transferring disturbances such as vibrations to the robot-patient interaction. In this paper, we investigate bending stiffness variability in dependence on the length of a flexible supply channel. Our experiments have shown that bending stiffness in fact decreases with the length of the supply channel but the highest stiffness connection is not sufficiently rigid and we plan to implement a more rigid connection to allow precise path following during insertion.