Registration of a statistical model to intraoperative ultrasound for scaphoid screw fixation

Anatomy-aware computed tomography-to-ultrasound spine registration

BACKGROUND

Ultrasound (US) has demonstrated to be an effective guidance technique for lumbar spine injections, enabling precise needle placement without exposing the surgeon or the patient to ionizing radiation. However, noise and acoustic shadowing artifacts make US data interpretation challenging. To mitigate these problems, many authors suggested using computed tomography (CT)-to-US registration to align the spine in pre-operative CT to intra-operative US data, thus providing localization of spinal landmarks.

PURPOSE

In this paper, we propose a deep learning (DL) pipeline for CT-to-US registration and address the problem of a need for annotated medical data for network training. Firstly, we design a data generation method to generate paired CT-US data where the spine is deformed in a physically consistent manner. Secondly, we train a point cloud (PC) registration network using anatomy-aware losses to enforce anatomically consistent predictions.

METHODS

Our proposed pipeline relies on training the network on realistic generated data. In our data generation method, we model the properties of the joints and disks between vertebrae based on biomechanical measurements in previous studies. We simulate the supine and prone position deformation by applying forces on the spine models. We choose the spine models from 35 patients in VerSe dataset. Each spine is deformed 10 times to create a noise-free data with ground-truth segmentation at hand. In our experiments, we use one-leave-out cross-validation strategy to measure the performance and the stability of the proposed method. For each experiment, we choose generated PCs from three spines as the test set. From the remaining, data from 3 spines act as the validation set and we use the rest of the data for training the algorithm. To train our network, we introduce anatomy-aware losses and constraints on the movement to match the physics of the spine, namely, rigidity loss and bio-mechanical loss. We define rigidity loss based on the fact that each vertebra can only transform rigidly while the disks and the surrounding tissue are deformable. Second, by using bio-mechanical loss we stop the network from inferring extreme movements by penalizing the force needed to get to a certain pose.

RESULTS

To validate the effectiveness of our fully automated data generation pipeline, we qualitatively assess the fidelity of the generated data. This assessment involves verifying the realism of the spinal deformation and subsequently confirming the plausibility of the simulated ultrasound images. Next, we demonstrate that the introduction of the anatomy-aware losses brings us closer to state-of-the-art (SOTA) and yields a reduction of 0.25 mm in terms of target registration error (TRE) compared to using only mean squared error (MSE) loss on the generated dataset. Furthermore, by using the proposed losses, the rigidity loss in inference decreases which shows that the inferred deformation respects the rigidity of the vertebrae and only introduces deformations in the soft tissue area to compensate the difference to the target PC. We also show that our results are close to the SOTA for the simulated US dataset with TRE of 3.89 mm and 3.63 mm for the proposed method and SOTA respectively. In addition, we show that our method is more robust against errors in the initialization in comparison to SOTA and significantly achieves better results (TRE of 4.88 mm compared to 5.66 mm) in this experiment.

CONCLUSIONS

In conclusion, we present a pipeline for spine CT-to-US registration and explore the potential benefits of utilizing anatomy-aware losses to enhance registration results. Additionally, we propose a fully automatic method to synthesize paired CT-US data with physically consistent deformations, which offers the opportunity to generate extensive datasets for network training. The generated dataset and the source code for data generation and registration pipeline can be accessed via https://github.com/mfazampour/medphys_ct_us_registration.

Ultrasound-Based Registration for the Computer-Assisted Navigated Percutaneous Scaphoid Fixation

An ultrasound (US)-based computer-assisted approach has the potential to improve the accuracy and precision of screw placement for the percutaneous fixation of scaphoid fractures and also reduce the radiation dose for patient and clinical staff. Therefore, a surgical plan based on preoperative diagnostic computed tomography (CT) is registered with intraoperative US images, enabling a navigated percutaneous fracture fixation. However, approaches published so far rely on semimanual methods for intraoperative registration and are limited by long computation times. To address these challenges, we propose the employment of deep learning-based methods for US segmentation and registration in order to achieve a fast and fully automated yet robust registration process. For validation of the proposed US-based approach, we first provide a comparison of methods for segmentation and registration, assess their contribution to the overall error throughout our pipeline, and, finally, evaluate navigated screw placement in an in vitro study on 3-D printed carpal phantoms. Successful screw placement has been achieved for all ten screws, with deviations from the planned axis of 1.0 ± 0.6 and 0.7 ± 0.3 mm at the distal and proximal pole, respectively. The complete automation and total duration of about 12 s also allow seamless integration of our approach into the surgical workflow.

Knee Bone Models From Ultrasound

The number of total knee arthroplasties performed worldwide is on the rise. Patient-specific planning and implants may improve surgical outcomes but require 3-D models of the bones involved. Ultrasound (US) may become a cheap and nonharmful imaging modality if the shortcomings of segmentation techniques in terms of automation, accuracy, and robustness are overcome; furthermore, any kind of US-based bone reconstruction must involve some kind of model completion to handle occluded areas, for example, the frontal femur. A fully automatic and robust processing pipeline is proposed, generating full bone models from 3-D freehand US scanning. A convolutional neural network (CNN) is combined with a statistical shape model (SSM) to segment and extrapolate the bone surface. We evaluate the method in vivo on ten subjects, comparing the US-based model to a magnetic resonance imaging (MRI) reference. The partial freehand 3-D record of the femur and tibia bones deviate by 0.7-0.8 mm from the MRI reference. After completion, the full bone model shows an average submillimetric error in the case of the femur and 1.24 mm in the case of the tibia. Processing of the images is performed in real time, and the final model fitting step is computed in less than a minute. It took an average of 22 min for a full record per subject.

Computer-assisted Surgery for Scaphoid Fracture

The computer-assisted surgery (CAS) has significantly improved the accuracy, reliability and outcomes of traumatic, spinal, nerve surgery and many other operations with a less invasive way. The application of CAS for scaphoid fractures remains experimental. The related studies are scanty and most of them are cadaver researches. Some intrinsic defects from the registration procedure, scan and immobilization of limbs may inevitably result in deviations. Some deviations become more obvious with operations of small bones (such as scaphoid) although they are acceptable for spine and other orthopedic surgeries. We reviewed the current literatures on the applications of CAS for scaphoid operation and summarized technical principles, scan and registration methods, immobilization of limbs and their outcomes. On the basis of the data, we analyzed the limitations of this technique and envisioned its future development.

[Ultrasound-guided percutaneous Herbert screw for the treatment of fresh nondisplaced carpal scaphoid fracture]

Objective

To explore the effectiveness of ultrasound-guided percutaneous Herbert screw for the treatment of fresh nondisplaced carpal scaphoid fracture.

Methods

Between May 2013 and August 2015, 15 patients with fresh nondisplaced carpal scaphoid fractures (Krimmer type A2) were treated with ultrasound-guided Herbert screw fixation. There were 12 males and 3 females with an average age of 33.4 years (range, 18-51 years). The causes of injury included 9 cases of falls, 3 cases of training injuries, and 3 cases of machine injuries. The interval from injury to surgery was 2-15 days (mean, 5 days). No other complication was found. The operation time, intraoperative blood loss, intraoperative fluoroscopy times, and the fracture healing time were recorded. The wrist function was assessed by the modified Mayo wrist score standard.

Results

The operation time was 28-53 minutes (mean, 33.9 minutes); the intraoperative blood loss was 5-30 mL (mean, 10.5 mL); the intraoperative fluoroscopy was 2-6 times (mean, 2.6 times). All 15 patients were followed up 6-18 months (mean, 10.5 months). One patient developed pain and soreness in the skin of the nail entrance, and gradually relieved after fumigation. No complication such as infection occurred. All fractures healed clinically, and the healing time was 8-16 weeks (mean, 11.6 weeks). At last follow-up, the modified Mayo wrist score was 76-99 (mean, 92.5). Among them, 12 cases were excellent, 2 cases were good, and 1 case was fair, and the excellent and good rate was 93.3%.

Conclusion

Ultrasound-guided fixation with Herbert screw is a reliable treatment method for fresh nondisplaced carpal scaphoid fractures with small invasion, less bleeding, and small radiation damage.

Three-Dimensional Registration of Freehand-Tracked Ultrasound to CT Images of the Talocrural Joint

A rigid surface⁻volume registration scheme is presented in this study to register computed tomography (CT) and free-hand tracked ultrasound (US) images of the talocrural joint. Prior to registration, bone surfaces expected to be visible in US are extracted from the CT volume and bone contours in 2D US data are enhanced based on monogenic signal representation of 2D US images. A 3D monogenic signal data is reconstructed from the 2D data using the position of the US probe recorded with an optical tracking system. When registering the surface extracted from the CT scan to the monogenic signal feature volume, six transformation parameters are estimated so as to optimize the sum of monogenic signal features over the transformed surface. The robustness of the registration algorithm was tested on a dataset collected from 12 cadaveric ankles. The proposed method was used in a clinical case study to investigate the potential of US imaging for pre-operative planning of arthroscopic access to talar (osteo)chondral defects (OCDs). The results suggest that registrations with a registration error of 2 mm and less is achievable, and US has the potential to be used in assessment of an OCD' arthroscopic accessibility, given the fact that 51% of the talar surface could be visualized.