The synergy of synchrotron imaging and convolutional neural networks towards the detection of human micro-scale bone architecture and damage

The growing health and economic burden of bone fractures, their intricate multiscale features and the existing knowledge gaps in the comprehension of micro-scale bone damage occurrence make fracture diagnosis a challenging issue. In this scenario, deep-learning and artificial intelligence embody the new frontier of healthcare system, by overcoming the subjectivity of clinicians in the analysis of medical images. However, the preliminary attempts in exploiting the power of machine learning algorithms such as neural networks are still limited to bone macro-scale, while there is an evident lack in their application to smaller scales, where damage starts nucleating. Currently, speculations at the micro-scale are only feasible with the aid of high-resolution imaging techniques, that are particularly time consuming in terms of output images analysis. In this context, this works aims at combining the visualization of the micro-crack propagation mechanism with the promising application of convolutional neural networks. The implemented artificial intelligence tool is based for the first time on a large number of human synchrotron images coming from healthy and osteoporotic femoral heads tested under micro-compression. The designed convolutional neural networks are able to automatically detect lacunae and micro-cracks at different compression levels with high accuracy levels; indeed, with the baseline setup, networks achieve more than 0.99 level of accuracy for both cracks and lacunae, and more than 0.87 of the meanIoU adopted as validation metric. This approach is particularly encouraging for the development of powerful recognition system to comprehend bone micro-damage initiation and propagation, paving the way to the application of machine learning studies to bone micromechanics. This could be additionally crucial for future patient specific micro-scale observations to be related to the clinical practice.

The impact of fall-related loading rate on the formation of micro-damage in human cortical bone fracture

The quest for better predictive tools as well as new preventative and therapeutic measures for bone fragility and fracture has highlighted the need for greater mechanistic understanding of the bone fracture process. Cortical bone, the major load bearing part of the bone, employs different toughening mechanisms to either inhibit or slow down crack growth which leads to fracture. Among these toughening mechanisms, is the formation of a micro-damage process zone (MDPZ) around the region of the propagating crack. Investigations into the MDPZ to date have primarily been based on quasi-static or cyclic loading rate experiments which do not necessarily replicate physiological fracture rates. Consequently, the impact of fall-related loading rates on the formation of the micro-damage process zone was investigated comparing these to quasi-static loading rate equivalents. The size of MDPZ was found to be 42% smaller in the high-rate group compared to the quasi-static rate group. The smaller MDPZ size was associated with a brittle, unstable fracture behaviour and an overall smaller fracture resistance measure (J_max). This result points to the possibility of a strain rate hardening mechanism at the heart of micro-damage formation, which is hampered under high loading rates, resulting in lower overall fracture resistance.

The micro-damage process zone during transverse cortical bone fracture: No ears at crack growth initiation

OBJECTIVE

Apply high-resolution benchtop micro-computed tomography (micro-CT) to gain greater understanding and knowledge of the formation of the micro-damage process zone formed during traverse fracture of cortical bone.

METHODS

Bovine cortical bone was cut into single edge notch (bending) fracture testing specimens with the crack on the transverse plane and oriented to grow in the circumferential direction. We used a multi-specimen technique and deformed the specimens to various individual secant modulus loss levels (P-values) up to and including maximum load (Pmax). Next, the specimens were infiltrated with a BaSO₄ precipitation stain and scanned at 3.57-μm isotropic voxel size using a benchtop high resolution-micro-CT. Measurements of the micro-damage process zone volume, width and height were made. These were compared with the simple Irwin's process zone model and with finite element models. Electron and confocal microscopy confirmed the formation of BaSO₄ precipitate in micro-cracks and other porosity, and an interesting novel mechanism similar to tunneling.

RESULTS

Measurable micro-damage was detected at low P values and the volume of the process zone increased according to a second order polynomial trend. Both width and height grew linearly up to Pmax, at which point the process zone cross-section (perpendicular to the plane of the crack) was almost circular on average with a radius of approximately 550µm (approximately one quarter of the unbroken ligament thickness) and corresponding to the shape expected for a biological composite under plane stress conditions.

CONCLUSION

This study reports details of the micro-damage fracture process zone previously unreported for cortical bone. High-resolution micro-CT enables 3D visualization and measurement of the process zone and confirmation that the crack front edge and process zone are affected by microstructure. It is clear that the process zone for the specimens studied grows to be meaningfully large, confirming the need for the J-integral approach and it does not achieve steady state at Pmax in most specimens. With further development, this approach may become valuable towards better understanding the role of the process zone in cortical bone fracture and the effects of relevant modifications towards changes in fracture toughness in a cost effective way.

Application of digital volume correlation to study the efficacy of prophylactic vertebral augmentation

BACKGROUND

Prophylactic augmentation is meant to reinforce the vertebral body, but in some cases it is suspected to actually weaken it. Past studies only investigated structural failure and the surface strain distribution. To elucidate the failure mechanism of the augmented vertebra, more information is needed about the internal strain distribution. This study aims to measure, for the first time, the full-field three-dimensional strain distribution inside augmented vertebrae in the elastic regime and to failure.

METHODS

Eight porcine vertebrae were prophylactically-augmented using two augmentation materials. They were scanned with a micro-computed tomography scanner (38.8μm voxel resolution) while undeformed, and loaded at 5%, 10%, and 15% compressions. Internal strains (axial, antero-posterior and lateral-lateral components) were computed using digital volume correlation.

FINDINGS

For both augmentation materials, the highest strains were measured in the regions adjacent to the injected cement mass, whereas the cement-interdigitated-bone was less strained. While this was already visible in the elastic regime (5%), it was a predictor of the localization of failure, which became visible at higher degrees of compression (10% and 15%), when failure propagated across the trabecular bone. Localization of high strains and failure was consistent between specimens, but different between the cement types.

INTERPRETATION

This study indicated the potential of digital volume correlation in measuring the internal strain (elastic regime) and failure in augmented vertebrae. While the cement-interdigitated region becomes stiffer (less strained), the adjacent non-augmented trabecular bone is affected by the stress concentration induced by the cement mass. This approach can help establish better criteria to improve vertebroplasty.

μFE models can represent microdamaged regions of healthy and metastatically involved whole vertebrae identified through histology and contrast enhanced μCT imaging

Micro-damage formation within the skeleton is an important stimulant for bone remodeling, however abnormal build-up of micro-damage can lead to skeletal fragility. In this study, µCT imaging based micro finite element (μFE) models were used to evaluate tissue level damage criteria in whole healthy and metastatically-involved vertebrae. T13-L2 spinal segments were excised from osteolytic (n=3) and healthy (n=3) female athymic rnu/rnu rats. Osteolytic metastasis was generated by intercardiac injection of HeLa cancer cells. Micro-mechanical axial loading was applied to the spinal motion segments under μCT imaging. Vertebral samples underwent BaSO4 staining and sequential calcein/fuchsin staining to identify load induced micro-damage. μCT imaging was used generate specimen specific μFE models of the healthy and osteolytic whole rat vertebrae. Model boundary conditions were generated through deformable image registration of loaded and unloaded scans. Elevated stresses and strains were detected in regions of micro-damage identified through histological and BaSO4 staining within healthy and osteolytic vertebral models, as compared to undamaged regions. Additionally, damaged regions of metastatic vertebrae experienced significantly higher local stresses and strains than those in the damaged regions of healthy specimens. Areas identified by BaSO4 staining, however, yielded lower levels of stress and strain in damaged and undamaged regions of healthy and metastatic vertebrae as compared to fuschin staining. The multimodal (experimental, image-based and computational) techniques used in this study demonstrated the ability of local stresses and strains computed through µFE analysis to identify trabecular micro-damage, that can be applied to biomechanical analyses of healthy and diseased whole bones.