Currently implemented accuracy metrics in open-source libraries for segmentation by supervised machine learning are typically one-dimensional scores [1]. While extremely relevant to evaluate applicability in clinics, anatomical location of segmentation errors is often neglected.

This study aims to include the three-dimensional (3D) spatial information in the development of a novel framework for segmentation accuracy evaluation and comparison between different methods.

Predicted and ground truth (manually segmented) segmentation masks are meshed into 3D surfaces. A template mesh of the same anatomical structure is then registered to all ground truth 3D surfaces. This ensures all surface points on the ground truth meshes to be in the same anatomically homologous order. Next, point-wise surface deviations between the registered ground truth mesh and the meshed segmentation prediction are calculated and allow for color plotting of point-wise descriptive statistics. Statistical parametric mapping includes point-wise false discovery rate (FDR) adjusted p-values (also referred to as q-values).

The framework reads volumetric image data containing the segmentation masks of both ground truth and segmentation prediction. 3D color plots containing descriptive statistics (mean absolute value, maximal value,…) on point-wise segmentation errors are rendered. As an example, we compared segmentation results of nnUNet [2], UNet++ [3] and UNETR [4] by visualizing the mean absolute error (surface deviation from ground truth) as a color plot on the 3D model of bone and cartilage of the mean distal femur.

A novel framework to evaluate segmentation accuracy is presented. Output includes anatomical information on the segmentation errors, as well as point-wise comparative statistics on different segmentation algorithms. Clearly, this allows for a better informed decision-making process when selecting the best algorithm for a specific clinical application.

We have investigated the errors in the identification of the transepicondylar axis and the anteroposterior axis between a minimally-invasive and a conventional approach in four fresh-frozen cadaver knees. The errors in aligning the femoral prosthesis were compared with the reference transepicondylar axis as established by CT.

The error in the identification of the transepicondylar axis was significantly higher in the minimal approach (4.5° of internal rotation, sd 4) than in the conventional approach (3° of internal rotation, sd 4; p < 0.001). The errors in identifying the anteroposterior axis in the two approaches were 0° (sd 5) and 1.8° (sd 5) of internal rotation, respectively (p < 0.001).

Conventional marker based optical motion capture (mocap) methods for estimating the position and orientation (pose) of anatomical segments use assumptions that anatomical segments are rigid bodies and the position of tracking markers is invariant relative to bones. Soft tissue artefact (STA) is the error in pose estimation due to markers secured to soft tissue that moves relative to bones. STA is a major source of pose estimation error and is most prevalent when markers are placed over joints. Mocap and bi-plane videoradiography data were recorded synchronously while three individuals walked on a treadmill. For all three, pose of the thigh and shank, and movement of markers relative to the bones, were determined from the videoradiography data (DSX, C-Motion). Independently, pose of thighs and shanks was estimated using mocap data (Visual3D, C-Motion). Our measures of error in the mocap pose estimation were the relative thigh and shank translations. X-ray data from two subjects were used to generate a regression model for the antero/posterior movement of the lateral knee marker against internal/external hip rotation. The mocap translation errors of the third subject, attributed to STA of the knee marker, were 15.6mm and 32.0mm respectively. The pose of the third subject was then estimated using a probabilistic algorithm incorporating our regression model. Mocap translation errors were reduced to 10.6mm (thigh) and 4.4mm (shank). The results from these data suggest that errors in pose estimation due to STA may possibly be reduced via the application of algorithms based on probabilistic inference to mocap data.

Summary Statement

The current study introduced the effects of projection errors on ankle morphological measurements using CT-based simulated radiographs by correlation analysis between 2D/3D dimensions and reliability analysis with randomised perturbations while measuring planar parameters on radiographs.

We have evaluated in vitro the accuracy of percutaneous and ultrasound registration as measured in terms of errors in rotation and version relative to the bony anterior pelvic plane in computer-assisted total hip replacement, and analysed the intra- and inter-observer reliability of manual or ultrasound registration. Four clinicians were asked to perform registration of the landmarks of the anterior pelvic plane on two cadavers. Registration was performed under four different conditions of acquisition. Errors in rotation were not significant. Version errors were significant with percutaneous methods (16.2°; p < 0.001 and 19.25° with surgical draping; p < 0.001), but not with the ultrasound acquisition (6.2°, p = 0.13). Intra-observer repeatability was achieved for all the methods. Inter-observer analysis showed acceptable agreement in the sagittal but not in the frontal plane. Ultrasound acquisition of the anterior pelvic plane was more reliable in vitro than the cutaneous digitisation currently used

Background and Aims. Many orthopaedic patients admitted to hospital who require urgent surgery are also on Warfarin. Patients with an INR>2 have an increased risk of bleeding complications during surgery; however delay to surgery due to a high INR has both clinical and financial implications. This audit evaluates whether the appropriate management for correction of INR is employed as per local guidelines and, if not, whether this results in significant delay to operative treatment. Methods. A retrospective and prospective audit was performed analysing all Orthopaedic trauma admissions admitted to University Hospital Aintree in a 5 month period. Only those solely on warfarin, who were not acutely bleeding and required surgery in <24 hours were included. Results. A sample of 17 patients was obtained of which only 35.3% had correct INR reversal as per local guidelines. Errors that occurred included 81.1% not being given further Vitamin K at reassessment, 18.2% given too much Vitamin K, 9.1% given too little Vitamin K, 18.2% given Octaplex incorrectly and 9.1% not given Octaplex when indicated. Only 1 patient had a delay to their surgery directly resulting from incorrect INR reversal (total time to surgery − 33 hours 45 minutes). Conclusion. Better education for clinical staff on the local policy for INR reversal in patients requiring urgent orthopaedic surgery is needed. Local policy guidelines have since been redesigned in light of these results and a completion audit cycle has been performed showing significant improvement with 50% of the patient sample given correct INR reversal

Anterior-posterior (AP) x-rays are routinely taken following total hip replacement to assess placement and orientation of implanted components. Pelvic orientation at the time of an AP x-ray can influence projected implant orientation. 1. However, the extent of pelvic orientation varies between patients. 2. Without compensation for patient specific pelvic orientation, misleading measurements for implant orientation may be obtained. These measurements are used as indicators for post-operative dislocation stability and range of motion. Errors in which could result in differences between expectations and the true outcome achieved. The aim of this research was to develop a tool that could be utilised to determine pelvic orientation from an AP x-ray. An algorithm based on comparing projections of a statistical shape model of the pelvis (n=20) with the target X-ray was developed in MATLAB. For each iteration, the average shape was adjusted, rotated (to account for patient-specific pelvic orientation), projected onto a 2D plane, and the simulated outline determined. With respect to rotation, the pelvis was allowed to rotate about its transverse axis (pelvic flexion/extension) and anterior-posterior axis (pelvic adduction/abduction). Minimum root mean square error between the outline of the pelvis from the X-ray and the projected shape model outline was used to select final values for flexion and adduction. To test the algorithm, virtual X-rays (n=6) of different pelvis in known orientations were created using the algorithm described by Freud et al. 3. The true pelvic orientation for each case was randomly generated. Angular error was defined as the difference between the true pelvic orientation and that selected by the algorithm. Initial testing has exhibited similar accuracy in determining true pelvic flexion (x̄error = 2.74°, σerror=±2.21°) and true pelvic adduction (x̄error = 2.38°, σerror=±1.76°). For both pelvic flexion and adduction the maximum angular error observed was 5.62°. The minimum angular error for pelvic flexion was 0.37°, whilst for pelvic adduction it was 1.08°. Although the algorithm is still under development, the low mean, maximum, and standard deviations of error from initial testing indicate the approach is promising. Ongoing work will involve the use of additional landmarks for registration and training shapes to improve the shape model. This tool will allow surgeons to more accurately determine true acetabular orientation relative to the pelvis without the use of additional x-ray views or CT scans. In turn, this will help improve diagnoses of post-operative range of motion and dislocation stability

The Adora RSA (NRT, Denmark) is a new stereo X-ray system custom built for Radeostereometry. Images are acquired using CXDI50C digital detectors (Canon, Netherlands). Analysis software was written locally to detect both Tantalum markers and the spherical head of the hip implant, and for RSA reconstruction and kinematic analysis. To assess geometric reproducibility, a planar grid phantom was constructed with 1400 2mm markers in a grid pattern over a 350 by 430 mm glass plate. Additionally 25 tantalum markers of each diameter 1.0, 0.8 and 0.5 mm were added within a 120mm square of the grid. The phantom was imaged repeatedly with translation and rotation over the detector. For small phantom movements of up to 10mm over the detector, very small measurement errors were observed of median 2 microns, maximum 6 microns. For larger movements, the errors increased to median 5 microns and maximum 50 microns. Errors also increased with decreasing exposure. For RSA validation, an acetabular PE cup was cemented to a Sawbone pelvis. Tantalum markers were inserted into the pelvis (10), cement (4), and cup (10). A 28mm metal head was fixed to the cup. The phantom was imaged repeatedly without movement, then moved in translation (up to 100 mm) and rotation (all axes, up to 45 degrees), and with full X-ray repositioning. Precision errors were calculated on the assumption of no relative movement between components. Results are given for repositioning movement categorised as none, small (less than 25mm or 15 degrees), medium (less than 50mm or 30 degrees), and large. For the head, the mean total point motion error was 4, 10, 14 and 24 micrometers. Mean error of segment fitting was less than 60 microns with no markers rejected from the composite segment of 24 markers. Cup migration total translation error was 10, 16, 24, and 35 micrometers with rotation errors less than 0.05 degrees. Observed RSA errors were small, increasing with phantom movement. This is consistent with the geometric uniformity tests. X-ray exposure and tissue thickness were also identified as factors in precision. We conclude this system has excellent precision for Radiostereometry

Summary Statement. Incorrect pedicle screw placement can lead to neurological complications. Practice outside the operating room on realistic bone models, with force feedback, could improve safety. Pedicle forces in cadaveric specimens are reported, to support development of a training tool for residents. Introduction. Inserting screws into the vertebral pedicles is a challenging step in spinal fusion and scoliosis surgeries. Errors in placement can lead to neurological complications and poor mechanical fixation. The more experienced the surgeon, the better the accuracy of the screw placement. A physical training system would provide orthopaedic residents with the feel of performing pedicle cannulation before operating on a patient. The proposed system consists of realistic bone models mimicking the geometry and material properties of typical patients, coupled with a force feedback probe. The purpose of the present study was to determine the forces encountered during pedicle probing to aid in the development of this training system. Methods. We performed two separate investigations. In the first study, 15 participants (9 expert surgeons, 3 fellows, 3 residents) were asked to press a standard pedicle awl three times onto a mechanical scale, blinded to the force, demonstrating what force they would apply during safe pedicle cannulation and during unsafe cortical breach. In the second study, three experienced surgeons used a standard pedicle awl fitted with a one-degree of freedom load cell to probe selected thoracolumbar vertebrae of eight cadaveric specimens to measure the forces required during pedicle cannulation and deliberate breaching, in randomised order. A total of 42 pedicles were tested. Results. Both studies had wide variations in the results, but were in general agreement. Cannulation (safe) forces averaged approximately 90 N (20 lb) whereas breach (unsafe) forces averaged approximately 135–155 N (30–35 lb). The lowest average forces in the cadaveric study were for pedicle cannulation, averaging 86 N (range, 23–125 N), which was significantly lower (p<0.001) than for anterior breach (135 N; range, 80–195 N); medial breach (149 N; range, 98–186 N) and lateral breach (157 N; range, 114–228 N). There were no significant differences among the breach forces (p>0.1). Cannulation forces were on average 59% of the breach forces (range, 19–84%) or conversely, breach forces were 70% higher than cannulation forces. Discussion. To our knowledge, axial force data have not previously been reported for pedicle cannulation and breaching. A large range of forces was measured, as is experienced clinically. Additional testing is planned with a six-degree-of-freedom load cell to determine all of the forces and moments involved in cannulation and breaching throughout the thoracolumbar spine. These results will inform the development of a realistic bone model as well as a breach prediction algorithm for a physical training system for spine surgery. The opportunity to learn and practice outside of the operating room, including learning from deliberate mistakes, should increase the confidence and comprehension of residents performing the procedure, enhance patient safety, reduce surgical time, and allow faster progression of learning inside the operating room

Objectives

Unicompartmental knee arthroplasty (UKA) is a demanding procedure, with tibial component subsidence or pain from high tibial strain being potential causes of revision. The optimal position in terms of load transfer has not been documented for lateral UKA. Our aim was to determine the effect of tibial component position on proximal tibial strain.