Micro-computed tomography-based highly automated 3D segmentation of the rat spine for quantitative analysis of metastatic disease

Treatment affects load to failure and microdamage accumulation in healthy and osteolytic rat vertebrae

Bone turnover and microdamage are impacted by the presence of skeletal metastases which can contribute to increased fracture risk. Treatments for metastatic disease may further impact bone quality. This exploratory study aimed to establish an initial understanding of microdamage accumulation and load to failure in healthy and osteolytic rat vertebrae following focal and systemic cancer treatment (docetaxel (DTX), stereotactic body radiotherapy (SBRT), or zoledronic acid (ZA)). Osteolytic spine metastases were developed in 6-week-old athymic female rats via intracardiac injection of HeLa human cervical cancer cells (day 0). Additional rats served as healthy controls. Rats were either untreated, received SBRT to the T10-L6 vertebrae on day 14 (15 Gy, two fractions), DTX on day 7 or 14, or ZA on day 7. Rats were euthanized on day 21. Tumor burden was assessed with bioluminescence images acquired on day 14 and 21, histology of the excised T11 and L5 vertebrae, and ex-vivo μCT images of the T13-L4. Microstructural parameters (bone volume/total volume, trabecular number, spacing, thickness, and bone mineral density) were measured from L2 vertebrae. Load to failure was measured with axial compressive loading of the L1-L3 motion segments. Microdamage accumulation was labeled in T13 vertebrae with BaSO4 staining and was visualized with high resolution μCT imaging. Microdamage volume fraction was defined as the ratio of BaSO4 to bone volume. DTX administered on day 7 reduced tumor growth significantly (p < 0.05). Microdamage accumulation was found to be increased by the presence of metastases but was reduced by all treatments with ZA showing the largest improvement in HeLa cell injected rats. Load to failure was decreased in untreated and SBRT HeLa cell injected rats compared to healthy controls (p < 0.01). There was a moderate negative correlation between load to failure and microdamage volume fraction in vertebrae from rats injected with HeLa cells (R = -0.35, p = 0.031). Strong correlations were also found between microstructural parameters and load to failure and microdamage accumulation. Several factors, including the presence of osteolytic lesions and use of cancer therapies, influence microdamage accumulation and load to failure in rat vertebrae. Understanding the impact of these treatments on fracture risk of metastatic vertebrae is important to improve management of patients with spinal metastases.

Man versus machine: Automatic pedicle screw planning using registration-based techniques compared with manual screw planning for thoracolumbar fusion surgeries

OBJECTIVE

This study evaluates the precision of a commercially available spine planning software in automatic spine labelling and screw-trajectory proposal.

METHODS

The software uses automatic segmentation and registration of the vertebra to generate screw proposals. 877 trajectories were compared. Four neurosurgeons assessed suggested trajectories, performed corrections, and manually planned pedicle screws. Additionally, automatic identification/labelling was evaluated.

RESULTS

Automatic labelling was correct in 89% of the cases. 92.9% of automatically planned trajectories were in accordance with G&R grade A + B. Automatic mode reduced the time spent planning screw trajectories by 7 s per screw to 20 s per vertebra. Manual mode yielded differences in screw-length between surgeons (largest distribution peak: 5 mm), automatic in contrast at 0 mm. The size of suggested pedicle screws was significantly smaller (largest peaks in difference between 0.5 and 3 mm) than the surgeon's choice.

CONCLUSION

Automatic identification of vertebrae works in most cases and suggested pedicle screw trajectories are acceptable. So far, it does not substitute for an experienced surgeon's assessment.

Machine learning for automated 3-dimensional segmentation of the spine and suggested placement of pedicle screws based on intraoperative cone-beam computer tomography

OBJECTIVE

The goal of this study was to develop and validate a system for automatic segmentation of the spine, pedicle identification, and screw path suggestion for use with an intraoperative 3D surgical navigation system.

METHODS

Cone-beam CT (CBCT) images of the spines of 21 cadavers were obtained. An automated model-based approach was used for segmentation. Using machine learning methodology, the algorithm was trained and validated on the image data sets. For measuring accuracy, surface area errors of the automatic segmentation were compared to the manually outlined reference surface on CBCT. To further test both technical and clinical accuracy, the algorithm was applied to a set of 20 clinical cases. The authors evaluated the system's accuracy in pedicle identification by measuring the distance between the user-defined midpoint of each pedicle and the automatically segmented midpoint. Finally, 2 independent surgeons performed a qualitative evaluation of the segmentation to judge whether it was adequate to guide surgical navigation and whether it would have resulted in a clinically acceptable pedicle screw placement.

RESULTS

The clinically relevant pedicle identification and automatic pedicle screw planning accuracy was 86.1%. By excluding patients with severe spinal deformities (i.e., Cobb angle > 75° and severe spinal degeneration) and previous surgeries, a success rate of 95.4% was achieved. The mean time (± SD) for automatic segmentation and screw planning in 5 vertebrae was 11 ± 4 seconds.

CONCLUSIONS

The technology investigated has the potential to aid surgeons in navigational planning and improve surgical navigation workflow while maintaining patient safety.

Elevated Microdamage Spatially Correlates with Stress in Metastatic Vertebrae

Metastasis of cancer to the spine impacts bone quality. This study aims to characterize vertebral microdamage secondary to metastatic disease considering the pattern of damage and its relationship to stress and strain under load. Osteolytic and mixed osteolytic/osteoblastic vertebral metastases were produced in athymic rats via HeLa cervical or canine Ace-1 prostate cancer cell inoculation, respectively. After 21 days, excised motion segments (T12-L2) were µCT scanned, stained with BaSO₄ and re-imaged. T13-L2 motion segments were loaded in axial compression to induce microdamage, re-stained and re-imaged. L1 (loaded) and T12 (unloaded) vertebrae were fixed, sample blocks cut, polished and BSE imaged. µFE models were generated of all L1 vertebrae with displacement boundary conditions applied based on the loaded µCT images. µCT stereological analysis, BSE analysis and µFE derived von Mises stress and principal strains were quantitatively compared (ANOVA), spatial correlations determined and patterns of microdamage assessed qualitatively. BaSO₄ identified microdamage was found to be spatially correlated with regions of high stress in µFEA. Load-induced microdamage was shown to be elevated in the presence of osteolytic and mixed metastatic disease, with diffuse, crossed hatched areas of microdamage present in addition to linear microdamage and microfractures in metastatic tissue, suggesting diminished bone quality.

A Mineralized Collagen-Polycaprolactone Composite Promotes Healing of a Porcine Mandibular Defect

A tissue engineering approach to address craniofacial defects requires a biomaterial that balances macro-scale mechanical stiffness and strength with the micron-scale features that promote cell expansion and tissue biosynthesis. Such criteria are often in opposition, leading to suboptimal mechanical competence or bioactivity. We report the use of a multiscale composite biomaterial that integrates a polycaprolactone (PCL) reinforcement structure with a mineralized collagen-glycosaminoglycan scaffold to circumvent conventional tradeoffs between mechanics and bioactivity. The composite promotes activation of the canonical bone morphogenetic protein 2 (BMP-2) pathway and subsequent mineralization of adipose-derived stem cells in the absence of supplemental BMP-2 or osteogenic media. We subsequently examined new bone infill in the acellular composite, scaffold alone, or PCL support in 10 mm dia. ramus mandibular defects in Yorkshire pigs. We report an analytical approach to quantify radial, angular, and depth bone infill from micro-computed tomography data. The collagen-PCL composite showed improved overall infill, and significantly increased radial and angular bone infill versus the PCL cage alone. Bone infill was further enhanced in the composite for defects that penetrated the medullary cavity, suggesting recruitment of marrow-derived cells. These results indicate a multiscale mineralized collagen-PCL composite offers strategic advantages for regenerative repair of craniofacial bone defects.

Automated Bone Segmentation and Surface Evaluation of a Small Animal Model of Post-Traumatic Osteoarthritis

MicroCT imaging allows for noninvasive microstructural evaluation of mineralized bone tissue, and is essential in studies of small animal models of bone and joint diseases. Automatic segmentation and evaluation of articular surfaces is challenging. Here, we present a novel method to create knee joint surface models, for the evaluation of PTOA-related joint changes in the rat using an atlas-based diffeomorphic registration to automatically isolate bone from surrounding tissues. As validation, two independent raters manually segment datasets and the resulting segmentations were compared to our novel automatic segmentation process. Data were evaluated using label map volumes, overlap metrics, Euclidean distance mapping, and a time trial. Intraclass correlation coefficients were calculated to compare methods, and were greater than 0.90. Total overlap, union overlap, and mean overlap were calculated to compare the automatic and manual methods and ranged from 0.85 to 0.99. A Euclidean distance comparison was also performed and showed no measurable difference between manual and automatic segmentations. Furthermore, our new method was 18 times faster than manual segmentation. Overall, this study describes a reliable, accurate, and automatic segmentation method for mineralized knee structures from microCT images, and will allow for efficient assessment of bony changes in small animal models of PTOA.

Effect of micro-computed tomography voxel size and segmentation method on trabecular bone microstructure measures in mice

Micro-computed tomography (μCT) is currently the gold standard for determining trabecular bone microstructure in small animal models. Numerous parameters associated with scanning and evaluation of μCT scans can strongly affect morphologic results obtained from bone samples. However, the effect of these parameters on specific trabecular bone outcomes is not well understood. This study investigated the effect of μCT scanning with nominal voxel sizes between 6–30 μm on trabecular bone outcomes quantified in mouse vertebral body trabecular bone. Additionally, two methods for determining a global segmentation threshold were compared: based on qualitative assessment of 2D images, or based on quantitative assessment of image histograms. It was found that nominal voxel size had a strong effect on several commonly reported trabecular bone parameters, in particular connectivity density, trabecular thickness, and bone tissue mineral density. Additionally, the two segmentation methods provided similar trabecular bone outcomes for scans with small nominal voxel sizes, but considerably different outcomes for scans with larger voxel sizes. The Qualitatively Selected segmentation method more consistently estimated trabecular bone volume fraction (BV/TV) and trabecular thickness across different voxel sizes, but the Histogram segmentation method more consistently estimated trabecular number, trabecular separation, and structure model index. Altogether, these results suggest that high-resolution scans be used whenever possible to provide the most accurate estimation of trabecular bone microstructure, and that the limitations of accurately determining trabecular bone outcomes should be considered when selecting scan parameters and making conclusions about inter-group variance or between-group differences in studies of trabecular bone microstructure in small animals.

•

This study investigated the effect of μCT scanning with voxel sizes between 6–30 μm on mouse trabecular bone measures.

•

Two commonly used segmentation methods for determining a global segmentation threshold were compared.

•

Voxel size strongly affected trabecular bone parameters including connectivity density and trabecular thickness.

•

The segmentation methods yielded similar outcomes for scans with small voxel sizes, but not scans with large voxel sizes.

•

Suggest that high-resolution scans should be used when possible to accurately estimate trabecular bone microstructure.

Reproducibility of compartmental subchondral bone morphometry in the mouse tibiofemoral joint

AIM

Evidence suggests that subchondral bone can be used as a predictor for the onset of osteoarthritis. As such, there is a need to accurately and reproducibly quantify subchondral bone in areas where osteoarthritis develops. In this paper, we present a novel technique for the segmentation of subchondral bone in the tibiofemoral joint and assess the reproducibility of this method with multiple measures and users.

METHODS

The right hind leg of seven C57BL/6 mice were excised and imaged in μCT. The menisci and patella were manually segmented and the image data was Gaussian filtered and binarized. An in-house algorithm was used to generate cortical and epiphyseal volumes of interest and standard morphometric indices for bone were computed. The intraclass correlation coefficient (ICC), absolute precision error (PE(SD)), and precision error as a percentage of the coefficient of variation of the repeated measurements (PE(%CV)) were calculated for each index. Additionally, an inter-user study was performed using the same indices and statistics.

RESULTS

For repeated measures, ICC ranged from 0.869 (cortical bone volume fraction, femur) to 0.994 (degree of anisotropy, femur). Similarly, PE(%CV) ranged from 0.84% (cortical bone volume fraction, femur) to 5.11% (connectivity density, tibia). For repeated users, no effect was seen in the femur with a slight effect in the tibia.

CONCLUSIONS

A novel method for the automatic segmentation of cortical and epiphyseal bone is presented and is shown to be reproducible in C57BL/6 mice. This tool will allow for high-throughput studies of osteoarthritis in animal models.

Effect of the metastatic defect on the structural response and failure process of human vertebrae: an experimental study

BACKGROUND

Pathologic vertebral fractures are associated with intractable pain, loss of function and high morbidity in patients with metastatic spine disease. However, the failure mechanisms of vertebrae with lytic defects and the failed vertebrae's ability to retain load carrying capacity remain unclear.

METHODS

Eighteen human thoracic and lumbar vertebrae with simulated uncontained bone defects were tested under compression-bending loads to failure. Failure was defined as 50% reduction in vertebral body height. The vertebrae were allowed to recover under load and re-tested to failure using the initial criteria. Repeated measure ANOVA was used to test for changes in strength and stiffness parameters.

FINDINGS

Vertebral failure occurred via buckling and fracture of the cortex around the defect, followed by collapse of the defect region. Compared to the intact vertebrae, the failed vertebrae exhibited a significant loss in compressive strength (59%, p<0.001), stiffness (53%, p<0.05) and flexion (70%, p<0.01) strength. Significant reduction in anterior-posterior shear (strength (63%, p<0.01) and stiffness (67%, p<0.01)) and lateral bending strength (134%, p<0.05) were similarly recorded. In the intact vertebrae, apart from flexion strength (r(2)=0.63), both compressive and anterior-posterior shear strengths were weakly correlated with their stiffness parameters (r(2)=0.24 and r(2)=0.31). By contrast, in the failed vertebrae, these parameters were strongly correlated, (r(2)=0.91, r(2)=0.86, and r(2)=0.92, p<0.001 respectively).

INTERPRETATION

Failure of the vertebral cortex at the defect site dominated the initiation and progression of vertebral failure with the vertebrae failing via a consolidation process of the vertebral bone. Once failed, the vertebrae showed remarkable loss of load carrying capacity.

The benefits of photodynamic therapy on vertebral bone are maintained and enhanced by combination treatment with bisphosphonates and radiation therapy

Photodynamic therapy (PDT) has been shown to ablate tumors within vertebral bone and yield short-term improvements in vertebral architecture and biomechanical strength, in particular when combined with bisphosphonate (BP) treatment. Longer-term outcomes of PDT combined with current treatments for skeletal metastases are essential to understand its therapeutic potential. The objective of this study is to evaluate the response of vertebrae to PDT after a longer (6-week) time period, alone and combined with previous BP or radiation treatment (RT). Sixty-three female rnu/rnu rats were randomized to six treatment groups: untreated control, BP-only, RT-only, PDT-only, combined BP + PDT and combined RT + PDT. L2 vertebrae were structurally analyzed through µCT-based analysis, axial compressive load-to-failure testing and histological analysis of morphology, osteoid formation and osteoclast activity. Combined BP + PDT treatment yielded the largest improvements in bone architecture with combined RT + PDT treatment yielding similar findings, but of a lesser magnitude. Mechanically, ultimate force and stress were correlated to stereological parameters that demonstrated a positive structural effect from combinatory treatment. Increased osteoid formation was observed in both combination therapies without any significant differences in osteoclast activity. Overall, multimodality treatment demonstrated a sustained positive effect on vertebral structural integrity, motivating PDT as a minimally-invasive adjuvant treatment for spinal metastases.

Multimodal μCT/μMR based semiautomated segmentation of rat vertebrae affected by mixed osteolytic/osteoblastic metastases

PURPOSE

Multimodal microimaging in preclinical models is used to examine the effect of spinal metastases on bony structure; however, the evaluation of tumor burden and its effect on microstructure has thus far been mainly qualitative or semiquantitative. Quantitative analysis of multimodality imaging is a time consuming task, motivating automated methods. As such, this study aimed to develop a low complexity semiautomated multimodal μCT/μMR based approach to segment rat vertebral structure affected by mixed osteolytic/osteoblastic destruction.

METHODS

Mixed vertebral metastases were developed via intracardiac injection of Ace-1 canine prostate cancer cells in three 4-week-old rnu/rnu rats. μCT imaging (for high resolution bone visualization), T1-weighted μMR imaging (for bone registration), and T2-weighted μMR imaging (for osteolytic tumor visualization) were conducted on one L1, three L2, and one L3 vertebrae (excised). One sample (L1-L3) was processed for undecalcified histology and stained with Goldner's trichome. The μCT and μMR images were registered using a 3D rigid registration algorithm with a mutual information metric. The vertebral microarchitecture was segmented from the μCT images using atlas-based demons deformable registration, levelset curvature evolution, and intensity-based thresholding techniques. The μCT based segmentation contours of the whole vertebrae were used to mask the T2-weighted μMR images, from which the osteolytic tumor tissue was segmented (intensity-based thresholding).

RESULTS

Accurate registration of μCT and μMRI modalities yielded precise segmentation of whole vertebrae, trabecular centrums, individual trabeculae, and osteolytic tumor tissue. While the algorithm identified the osteoblastic tumor attached to the vertebral pereosteal surfaces, it was limited in segmenting osteoblastic tissue located within the trabecular centrums.

CONCLUSIONS

This semiautomated segmentation method yielded accurate registration of μCT and μMRI modalities with application to the development of mathematical models analyzing the mechanical stability of metastatically involved vertebrae and in preclinical applications evaluating new and existing treatment effects on tumor burden and skeletal microstructure.

Beyond radiation therapy: photodynamic therapy maintains structural integrity of irradiated healthy and metastatically involved vertebrae in a pre-clinical in vivo model

Spinal metastasis commonly occurs in advanced breast cancer. Treatment is often multimodal including radiation therapy (RT), bisphosphonates (BPs), and surgery, yet alternative minimally invasive local treatments are needed. Photodynamic therapy (PDT) has been shown to ablate tumor cells and enhance bone formation secondary to metastatic breast cancer, demonstrating potential as a treatment for spinal metastasis. Combined with previous BP treatment, bone formation was further enhanced by PDT. This study aimed to determine the effects of PDT in combination with previous RT on healthy and metastatically involved vertebrae. Forty-six athymic rats underwent RT (4 Gy on day-7), twenty-three of them were inoculated with MT-1 human breast cancer cells on day 0. Thirteen healthy and ten metastatically involved rats underwent PDT treatment on day 14. All rats were sacrificed on day 21. L2 vertebrae were analyzed using μCT imaging, mechanical testing, and histological methods. In healthy vertebrae, while modest increases in trabecular structure were found in RT + PDT compared to RT only, mechanical stability was negatively affected. The 4 Gy RT dose was found to ablate all tumor cells and prevent further vertebral metastasis. As such, in metastatically involved rats, no differences in stereological or mechanical properties were detected. RT + PDT and RT-only treatment resulted in greatly improved vertebral structural and mechanical properties versus untreated or PDT-only treatment in metastatically involved rats, due to early tumor destruction in RT-treated groups. Increased amounts of woven bone and osteoid volume were found in PDT-treated vertebrae. Further investigation is needed to understand if structural improvements seen in RT + PDT treatment can translate into longer-term improvements in strength to support the potential of PDT as a viable adjuvant treatment for spinal metastasis postradiation.

Atlas-based automatic mouse brain image segmentation revisited: model complexity vs. image registration

Although many atlas-based segmentation methods have been developed and validated for the human brain, limited work has been done for the mouse brain. This paper investigated roles of image registration and segmentation model complexity in the mouse brain segmentation. We employed four segmentation models [single atlas, multiatlas, simultaneous truth and performance level estimation (STAPLE) and Markov random field (MRF) via four different image registration algorithms (affine, B-spline free-form deformation (FFD), Demons and large deformation diffeomorphic metric mapping (LDDMM)] for delineating 19 structures from in vivo magnetic resonance microscopy images. We validated their accuracies against manual segmentation. Our results revealed that LDDMM outperformed Demons, FFD and affine in any of the segmentation models. Under the same registration, increasing segmentation model complexity from single atlas to multiatlas, STAPLE or MRF significantly improved the segmentation accuracy. Interestingly, the multiatlas-based segmentation using nonlinear registrations (FFD, Demons and LDDMM) had similar performance to their STAPLE counterparts, while they both outperformed their MRF counterparts. Furthermore, when the single-atlas affine segmentation was used as reference, the improvement due to nonlinear registrations (FFD, Demons and LDDMM) in the single-atlas segmentation model was greater than that due to increasing model complexity (multiatlas, STAPLE and MRF affine segmentation). Hence, we concluded that image registration plays a more crucial role in the atlas-based automatic mouse brain segmentation as compared to model complexity. Multiple atlases with LDDMM can best improve the segmentation accuracy in the mouse brain among all segmentation models tested in this study.