Orthopaedic Biomechanics

Deblurring clinical CT scans to restore thin bone structures

In clinical CT imaging, there are severe errors in geometry and material properties in thin bone structures, which are in large part due to blurring. As such we aimed to restore sub-mm bone geometry and intensity information in clinical CT images by post-reconstruction image processing. Knowing what the non-blurred profile of cortical bone should look like, we can estimate it based on the observed profile in low-resolution CT images to restore the 3D anatomy of the craniomaxillofacial skeleton (CMFS).

Finite Element modelling of the CMFS

We created a new pipeline to generate accurate Finite Element (FE) models of the CMFS with heterogenous application of bone modulus values based on deblurred CT image intensity. Due to the very thin nature of the bone, this used nodal based assignments and algorithms to deal with partial volume effects at surfaces.

Mechanical testing is performed on specimens instrumented with strain gages to validate our modelling. Under more physiologic loading conditions, these models are able to show important pathways of load transmission occurring through the complex areas of the zygoma and orbital rim.

Understanding the sensitivity to load and boundary conditions, we are employing new methods to better represent muscle loading. MR-based diffusion tensor imaging (DTI), in combination with anatomical data, can elucidate the fibres and fibre bundles associated with the muscles of mastication. We are concurrently developing models focused on representing loading post-fracture reconstruction towards optimizing hardware requirements for stable fixation.

A CAD-CAM pipeline for facial and skeletal reconstruction

In CMF reconstruction, the goal is to recreate the pre-traumatic appearance of the individual. Using pre-trauma 2D photographs, a pipeline of facial detection (finding faces in images), facial recognition (finding the same person’s face), facial landmarking (locating points of interest on the face) and face morph modelling (generating 3D face guided by landmarks) can be used to create a pre-trauma 3D face geometry. A ‘reverse-engineering’ of a forensics soft tissue model can then be used to determine the underlying skull shape from the 3D facial surface topography.

1. Pre-trauma photos
2. Patient provides reference photos to surgeon
3. Surgeon provides photos for analysis
4. Facial detection, facial recognition, 2D landmarking, 3D morph model
5. Pre-trauma 3D face geometry
6. Forensics’ soft-tissue thickness
7. Inferred skull
8. Nose skin & cartilage templates
9. Surgery

The pipeline which goes from photo to face model to underlying craniomaxillofacial skeleton (CMFS) must consider age, sex, BMI, ethnicity, and facial expression. Application for this is in both form and function considering re-establishment of appearance, structure and facial reanimation.

An application of the CAD-CAM pipeline is shown in nasal reconstruction following trauma or tumour resection. A forehead flap graft can be preoperatively planned based on a 3D facial reconstruction. Flattening software transforms the 3D shape into a physical template for the graft to direct intraoperative cutting of a skin flap.

Patient-specific CMFS implants made in the operating room

Calavera Surgical Design (a Sunnybrook start up company) has a process which creates unique patient-specific implants from standard commercial meshes that can be modified in the OR – enabling intraoperative sizing of the implants based on the actual resection. We have developed a new image analysis pipeline, including our deblurring algorithm, which has been integrated into Calavera’s clinical workflow, to reduce time and manual intervention in creating the designs.

In vivo study has demonstrated the healing of zygomatic fractures stabilized with Bone Tape without any adverse histologic reactions. Additional work is ongoing in optimizing the tape formulation and workflow leading to further in vivo preclinical testing. Commercialization of Bone Tape is being led by a new start-up company Cohesys.

Bone quality is important from mechanics and material perspectives when considering vertebral structural integrity. Quality is dependent on mechanical (material and structural) properties, including mineralization, composition, remodelling, connectivity and architecture. To understand how metastatic disease (osteolytic and osteoblastic) affects bone quality we developed preclinical nude rat models of Osteolytic (HeLa), Mixed (ACE-1) and Osteoblastic (ZR-75) metastases. These models highlighted changes in the properties of both the mineral (hydroxyapatite) and organic (collagen) phases of the metastatic bone tissue, which correlated to tissue level material behaviour and whole vertebral mechanical properties. Computationally we have developed algorithms to directly measure deformation and strain via registration of loaded and unloaded µCT scans.

1. Unloaded, Subvoxel Interpolation
2. Loaded, Subvoxel Interpolation
3. Multiresolution Deformable Registration
4. Deformation field
5. Strain

Load and boundary conditions from this experimental imaging data have been applied to continuum and micro finite element (µFE) models of healthy and metastatically involved vertebrae. Microdamage present in healthy and metastatically involved vertebrae can be identified histologically through fuchsin staining, chelating agents and barium sulphate staining combined with µCT imaging. For higher resolution information, back scatter electron microscopy can be performed on BaSO4 stained sections. µFE models of these vertebrae were able to represent the load induced microdamage seen in these images. Microdamage was found to be spatially correlated with regions of high stress and strain in our µFE models with clear microdamage thresholds identified. Yet mechanically induced microdamage thresholds were different in healthy vs metastatic bone suggesting tissue level changes in bone quality.

Recent work has gone beyond modelling of stress and strain, demonstrating the ability of non-linear FE models incorporating cohesive elements to represent post-yield modeling of crack initiation, propagation and failure. The impact of multimodal treatments on bone quality in the metastatic skeleton is an ongoing area of investigation.

Fracture risk prediction in the metastatic spine

Vertebral compression fracture risk prediction: A deep learning approach

U net Convolutional Neural Network (CNN) architectures have been shown to be fast and highly accurate for image segmentation problems. A combination of segmentation and classification information (fractured/not fractured) will ultimately be used to train and develop a CNN based approach for fracture risk prediction in the metastatic spine. Deeper networks using additional training data and systematic hyper-parameter optimization may further improve initial segmentation results. We also aim to combine CT and MR images using a latent model to enable multimodal vertebral segmentation with an ultimate goal of integrating this approach into a clinical scoring system.

Treatments of bone metastases are designed to decrease pain, increase structural stability, improve mobility and control tumour growth. Treatments include systemic drugs [Bisphosphonates (BPs), Rank-L inhibitors (i.e. Denosumab) and chemotherapeutic agents (e.g. docetaxel)], focal treatments (i.e. spine stereotactic body radiotherapy [SBRT] and radiofrequency ablation [RFA]) and structural stabilization (i.e. cement augmentation or instrumentation). Our research has primarily focused on focal minimally invasive treatment modalities, including consideration of these treatments within a more comprehensive multimodal approach to the treatment of skeletal metastases.

Photodynamic therapy

Used clinically in the treatment of macular degeneration and light accessible cancers such as melanoma, Photodynamic Therapy (PDT) involves systemic administration of a photosensitizer that accumulates preferentially in target tissue and becomes activated by non-thermal light at a specific wavelength.

Activation of the photosensitizer in the presence of molecular oxygen leads to generation of highly reactive singlet oxygen, which causes tumor cell toxicity and tissue necrosis. A minimally invasive approach allows direct delivery of laser light to spinal lesions. Initial testing demonstrated the ability of PDT to ablate tumour tissue and augment the surrounding bone.

A multimodal preclinical treatment study demonstrated the impact of PDT in conjunction with standard of care clinical treatments Bisphosphonates (BP) and Radiation therapy (RT). BP+PDT both destroyed the tumour and had the largest positive effect on bone, restoring bone volume fraction to non-tumour bearing control levels (with increased osteoid, woven bone, periosteal bone growth). An increase in bone volume was correlated to improved mechanical stability. Combined RT+PDT showed a similar effect on bone (lower magnitude).

Motivated by this promising preclinical work, a phase I clinical trial of PDT for spinal metastases has successfully been completed at Sunnybrook. Next steps will focus on treatment planning and running a Phase II clinical trial to quantify the direct impact of PDT on the vertebrae.

Radiofrequency Ablation (RFA)

Radiofrequency (RF) refers to the electromagnetic spectrum covering the frequencies from 3 Hz to 300 GHz. In the context of tumour destruction, RF ablation induces ionic modulation leading to controlled heat production & local ablation of targeted tissue.

In collaboration with Baylis Medical (Mississauga, ON) we developed a novel bipolar internally cooled RF probe for the ablation of spinal metastases. The probe was designed to overcome previous limitations with RF of small and incomplete ablated regions, charring, boiling, iatrogenic injury, and difficulty completing the RF circuit in bone. Large lesions were ablated with RFA without damage to the neighbouring critical structures. Osteoclasts and tumour cells were also shown to be very susceptible to RFA but destruction of osteocytes was more limited (desirable for bone health).

The device was brought to first in human testing and ultimately became an FDA clinically approved device, Osteocool. With the sale of the Osteocool system to Medtronic, there is now worldwide clinical use of this device. We are currently developing a treatment planning and navigation system for RFA of spinal metastases with an integrated approach to enable multimodal treatment with SBRT.

Pedicle screw instrumentation is the dominant surgical approach to achieve immediate stability of the spine. Transcortical screw insertion is an emerging technique of posterior spinal instrumentation with promising biomechanical properties; it requires a more medialized start point and follows an “up and out” trajectory. The objective of this work is to validate a freehand technique for insertion of transcortical screws.

A custom surgical simulator was used to determine appropriate freehand techniques of cortical screw insertion in the thoracic and lumbar spine. Specifically, case-specific volume rendering of non-pathologic and degenerative/scoliotic thoracic and lumbar spine CTs were used.

The start point and trajectory determination were modified from previously published methods by assessing screw fixation and purchase virtually using CT density data. Once identifiable and reproducible anatomic landmarks (Transverse Process, Pars interarticularis, midline) of the posterior spine were selected based on simulation, these methods were translated to the operating room. The vertebral level specific approach and landmarks were then checked against a Stryker based navigation system. In all cases, the freehand technique was deemed to place cortical screws ideally or safely.

The usage of transcortical screw fixation in the lumbar and thoracic spine may be a reasonable and expeditious alternative to pedicle screw fixation with the distinct advantage of limiting further lateral dissection, simpler and reproducible screw insertion start points and trajectories, and decreased risk of undesired screw breaches. This project demonstrates the power of virtual 3D interaction with medical imaging for surgical planning and technique generation.

As residency programs move to competency-based curricula, more authentic and accessible teaching tools are required to train the next generation of orthopaedic and neurosurgeons. Given the potential for neural complications, there exist significant barriers to residents and fellows obtaining adequate experience performing spinal surgery in the operating room. Virtual simulation tools are lacking for spinal surgery. The aim of this work is to develop an open-source 3D virtual simulator as a teaching tool to improve training in spinal decompression and pedicle screw insertion.

A custom stepwise simulator workflow was built using 3D Slicer; an open-source software development platform for medical image visualization and processing.

The procedural steps include import of patient-specific imaging, fusion of CT and MRI data, bone threshold-based segmentation, soft tissue segmentation, vertebral level identification, surgical planning, surgical field simulation, laminectomy, pedicle screw placement (start point, screw size, and trajectory), spinal decompression simulation, and evaluation. Bone and soft tissue resecting tools were developed by customizing manual 3D segmentation tools. Laminectomy simulation was enabled through bone and ligamentum flavum resection at the site of compression.

The completed workflow allows patient specific simulation and visualization of spinal procedures. Procedural accuracy, the design of resecting tools, and modeling of the impact of bone and ligament removal adequately encompass important challenges in spinal surgery. Screw-bone contact area can be evaluated quantitatively. Visualization of contact, decompression, tissue resection and positioning can be evaluated.

This software development project has resulted in a well-characterized open source tool for simulating spinal surgery. Future work will integrate the simulator within existing orthopaedic resident competency-based curriculum and fellowship training instruction.

Computed Tomography (CT) based navigation is rapidly becoming the standard of care in spinal surgery. While CT provides excellent bony detail, it is quite limited in its ability to discern soft tissues. Intra-procedural ultrasound (US) can provide real-time soft tissue discrimination allowing the delineation of tumor tissue, discs and neural elements, which is particularly useful for minimally invasive surgery. Challenges exist with application of US in the spine with respect to spinal level and orientation determination.

We have developed a spine-focused surgical workflow allowing for interchangeable US probes and navigation systems. Our technology uses the 3D Slicer open source platform for rapid development of image visualization and navigation solutions. We tested our technology with both Phillips and Telmed US systems, and a Medtronic StealthStation S7 for optical tracking.

The interface shows in real-time: 3D view of US plane with a rendering of the anatomy, US Image, corresponding CT image plane and fused image. Accuracy (mm) of the system was established by identifying features on the vertebral body in both modalities.

Our spine surgery workflow allows US-CT fusion intra-operatively, simultaneously displaying hard and soft tissue information, using various combinations of US systems. Accurate localization enables the utilization of navigation to guide soft tissue resection in places that are difficult to visualize (i.e. in thoracic discectomy or assessing the adequacy of a dorsal decompression during a cervical laminectomy for myelopathy).

The minimum important difference (MID) observed following spine surgery has varied drastically in the literature when considering PROMs. Here we are Investigating changes in patient’s MID over recovery from spinal surgery and how cognitive appraisal processes are implicated in the change trajectories.

Measures:

• Rand-36 Physical and Mental Component Scores (PCS and MCS)
• Oswestry Disability Index (ODI)
• Pain Numeric Rating Scale Items (at rest, with activity, back and leg)
• PROMIS pain impact
• Global Assessment of Change (GAC) item
• The QOL Appraisal Profile – Standards of comparison domain

Among patients with different trajectories after surgery, the relationship between PROM change and appraisal change differs by group. Better, Worse, and Bouncer patients are focusing on different standards of comparison over time.

Factors affecting minimal important difference:

• Who you are asking (better or worse)
• How disabled are they at baseline
• When you are asking (time from surgery)
• What the patient is comparing themselves to

By identifying whether and which appraisal processes influence PROM trajectories, the present work provides potential avenues for clinical intervention to improve spine surgery outcomes.

Challenges in intramedullary (IM) nailing of long bone fractures arise during the entry point step in which 2D fluoroscopic imaging is used to guide 3D alignment of a guidewire safely into the IM canal. Specifically, the guidewire alignment achieved under fluoroscopy in one plane may be lost while adjusting the alignment in the orthogonal plane. This impedes the surgical workflow requiring unpredictable repetition of surgical activities (e.g. repeated fluoroscopic repositioning) and leads to frustration in the operating room. Such challenges are particularly evident in large (muscular and obese) patients.

FAST (Femoral Antegrade Starting Tool) is an innovative surgical tool that facilitates entry point selection and insertion for IM nailing. FAST enables maintenance of guidewire (Kirschner (K) wire) anteroposterior (AP) alignment during lateral imaging and K-wire positioning in the sagittal plane addressing both localization and orientation challenges that arise from the use of 2D imaging to achieve 3D alignment.

In use for femoral fracture stabilization, the initial FAST device entry point placement is conducted freehand at the greater trochanter under fluoroscopic image guidance. The device location and orientation are adjusted until satisfactory AP alignment is obtained. The device is fixed onto the greater trochanter, locking the AP alignment. Under lateral fluoroscopic imaging, the K-wire is inserted into a channel of the FAST device best aligned with the medullary canal. The FAST arm is rotated in plane to achieve optimal lateral alignment. With alignment optimized in both planes, the K-wire can be safely drilled into the bone. This workflow has been shown in preclinical study to minimize time and radiation exposure from fluoroscopic imaging in femoral IM nailing. A first-in-human trial of FAST was successfully completed in 2019.

Positioning of the glenoid component is one of the most challenging steps in total shoulder arthroplasty. Prosthetic longevity as well as functional outcomes are considered highly dependent on accurate positioning. Currently, there are no adequate means to verify the position of the glenoid component intra-operatively which is a significant impediment to accurate positioning. To address this clinical need we have developed Bullseye, a novel intra-operative imaging system that utilizes a hand-held structured light sensor and computer vision algorithms to verify the 3D position of the glenoid vault guide pin prior to preparation of the glenoid for component implantation.

The canonical Wnt/β-Catenin pathway is a promising therapeutic target to stimulate bone growth. A molecular mechanism known to play a prominent role in skeletal development during embryogenesis, and believed to be recapitulated during fracture repair, the Wnt/β-Catenin pathway has been confirmed to influence the osteoblast lineage, stimulating mesenchymal progenitor precursors to differentiate into fully active, mature osteoblasts. Lithium is an integral component of psychotropic medicine but has also been linked to the Wnt/β-Catenin pathway and has been shown to positively influence bone biology acting as an anabolic agent to enhance fracture repair.

Through preclinical testing using a factorial design of experiments approach, we optimized the dose and duration of timing of lithium administration for fracture healing in healthy and osteoporotic femurs. It was found that the best treatment combination occurred at a low dose (20 mg/kgwt/day), late onset (7 days post-fracture) and a 2-week duration in healthy rats. Optimized lithium treatment yielded a 46 per cent increase in strength (maximum yield torque) at 4 weeks post-fracture over pooled controls. A similar 50 per cent increase in strength was seen with a further delayed administration of lithium (10 days post-fracture) in treating osteoporotic femurs at a six-week endpoint.

A Phase II Clinical Trial in patients (LiFT) is currently underway at Sunnybrook to determine if lithium can improve clinical long bone fracture healing. This research is also considering potential barriers to the uptake of lithium therapy for fracture healing from the perspective of both patients and care providers.

Additional preclinical work is underway examining the impact of the microbiome on fracture healing.

Fractures of the pelvis occur due to high-energy impacts (traumatic fractures) and the application of physiologic loads to bone deficient in mineral or elastic resistance (insufficiency fractures). Displaced fractures can lead to disabling post-traumatic arthritis; some fracture patterns induce worse prognosis than others. However, clinical decision-making regarding patient management can be very complex. A better understanding of pelvic and acetabular mechanics and changes in the bone’s mechanical strength could lead to improved diagnosis and treatment of insufficiency and traumatic pelvic and acetabular fractures.

Finite element (FE) analysis is a powerful tool that has been successful in predicting the strength of healthy and pathologic bones. Patient-specific FE modeling is able to represent bones, which are highly heterogeneous in their presentation and are difficult to represent with other analytical or parametrically developed techniques. We have developed and experimentally validated specimen-specific 3D FE models of the pelvis and analyzed them under physiological loading. Using these models, we have examined the sensitivity of pelvic strain to bone material properties, shape difference and loading scenarios.

Physical therapy is essential for the successful rehabilitation of common shoulder injuries and following shoulder surgery. Patients may receive some training and supervision for shoulder physiotherapy through private pay or private insurance, but they are typically responsible for performing most of their physiotherapy independently at home. It is unknown to what degree patients perform their home exercises and if the exercises are done correctly without supervision. Our team has recently developed a Smart Physiotherapy Activity Recognition System (SPARS) for tracking home shoulder physiotherapy exercises using sensors in a commercial smart watch and artificial intelligence (AI). SPARS has been successfully shown to track shoulder exercises in healthy adults in the laboratory setting. Further inquiry is required to establish the clinical effectiveness of this technology and investigate the potential individual and societal impacts of its use. A clinical study focused on both implementation and implications of adherence monitoring with AI in patients with rotator cuff pathology is planned to be carried out within the Working Conditions Program at the Holland Centre.

Open source code available

Seglearn is a python package for machine learning time series or sequences which was developed for this project. It provides an integrated pipeline for segmentation, feature extraction, feature processing, and final estimator. Seglearn provides a flexible approach to multivariate time series and related contextual (meta) data for classification, regression, and forecasting problems. Support and examples are provided for learning time series with classical machine learning and deep learning models. It is compatible with scikit-learn.

Waitlists for orthopaedic surgery, particularly hip and knee replacement are growing. Significant resources are required to run an operating room, making it one of the most valuable departments for hospitals and health care systems. Therefore, as resources – both physical and human – are limited across the public system in Canada, efficient management is crucial to the timely delivery of care. Assigning a specific date, time, operating room, surgeon, nurses, recovery bed and ward bed quickly becomes a daunting task for administrators due to the large volumes of cases. Our research challenges the tradition of tackling the highly complex orthopaedic surgery scheduling tasks using solely manual approaches based on surgeons’ estimates of operative time. Instead, we aim to design a comprehensive solution that leverages patient data, case volumes and patient-specific operative times to optimize the scheduling process in an automated way. Machine learning and mathematical optimization methods will be developed to enable both accurate predictions of operative time and optimal scheduling. In addition to financial benefits, optimized schedules would decrease patient waitlist times by facilitating more surgeries within the same amount of operating room time.

• Canadian Institutes of Health Research (CIHR)
• Canadian Orthopaedic Foundation
• National Science and Engineering Research Council (NSERC)
• Ontario Graduate Scholarships
• Varian Medical Systems
• William and Susanne Holland
• Workplace Safety and Insurance Board (WSIB)
• Wyss Medical Foundation

• AO Foundation
• Bayer Canada
• Canada Foundation for Innovation
• Canadian Breast Cancer Foundation
• Federal Economic Development Agency for Southern Ontario
• MITACS
• NVIDIA Corporation
• Ontario Centres of Excellence
• Ontario Innovation Trust
• Ontario Institute for Cancer Research
• Qatar National Research Fund

Industry collaborators

• Anchor Orthopaedics – view company website
• Baylis Medical – view company website
• Calavera Surgical Design – view company website
• Cohesys – view company website
• CURV
• Medtronic – view company website
• Synaptive Medical – view company website
• Thermal Technology Services Canada