In this protocol, the femur surface strains are estimated during fracture testing using the digital image correlation technique. The novelty of the method involves application of a high-contrast stochastic speckle pattern on the femur surface, carefully specified illumination, high speed video capture, and digital image correlation analysis for strain calculations.
This protocol describes the method using digital image correlation to estimate cortical strain from high speed video images of the cadaveric femoral surface obtained from mechanical testing. This optical method requires a texture of many contrasting fiduciary marks on a solid white background for accurate tracking of surface deformation as loading is applied to the specimen. Immediately prior to testing, the surface of interest in the camera view is painted with a water-based white primer and allowed to dry for several minutes. Then, a black paint is speckled carefully over the white background with special consideration for the even size and shape of the droplets. Illumination is carefully designed and set such that there is optimal contrast of these marks while minimizing reflections through the use of filters. Images were obtained through high speed video capture at up to 12,000 frames/s. The key images prior to and including the fracture event are extracted and deformations are estimated between successive frames in carefully sized interrogation windows over a specified region of interest. These deformations are then used to compute surface strain temporally during the fracture test. The strain data is very useful for identifying fracture initiation within the femur, and for eventual validation of proximal femur fracture strength models derived from Quantitative Computed Tomography-based Finite Element Analysis (QCT/FEA).
Digital Image Correlation (DIC) is an image post-processing method that is used in the current protocol to estimate the full field surface strain of cadaveric femoral test specimens from time-sequence images obtained during mechanical fracture tests. The technique was first developed and applied in experimental stress analysis in the 1980's and has experienced a rapid increase in use in recent years1,2,3. It has several key advantages over more traditional approaches of mounting strain gauges on a structure including increased spatial distribution of the strain field, finer gauge lengths through increased camera resolution, and avoiding issues with strain gauge glue adhesion or compliance. A major advantage of DIC for biological tissues, such as bone, is that it can be applied to irregular geometries comprising of highly heterogeneous material properties4,5. Its primary drawback over traditional strain acquisition methods is that it requires expensive high speed video cameras of sufficient resolution for the measurement of the region of interest to achieve sufficient spatial and temporal sampling to accurately estimate strain fields.
The primary application of the temporal strain fields obtained from bone fracture DIC analysis is to validate the strain estimations in QCT/FEA models of femoral strength5. Such validation is the focus of many orthopedic research groups which predominantly utilize remote measurements of force and displacement from load cells and displacement transducers6,7,8. In addition, post-fracture image analysis of the fracture pattern has been combined with these remote measurements as further means of model validation9. More recently, the DIC method was applied to validate an FEA model of fracture and crack propagation in the proximal femur10. By utilizing strain correlation between models and experiments, even more confidence in the validity of computational models of proximal femora will be obtained and further advance the QCT/FEA diagnostic method closer to clinical use.
This work explains a detailed protocol to incorporate the necessary steps for DIC analysis in fracture testing of proximal femora. The procedure included the bone preparation steps of spraying a white paint on the bone surface and then speckling black spots on the dried white surface of the bone, methods of obtaining images with sufficient spatial and temporal resolution using high speed video cameras, and the process and tools we used for computing strain fields from these images. We also explained several caveats that may affect the quality of the measurements.
All experiments were conducted with Institutional Review Board approval. The samples were obtained from anatomical research labs in collaboration.
1. Preparing Specimens for Testing
2. Speckling Process
3. Image Acquisition
4. Image Preparation
5. Finite Element Mesh Creation
6. Register the FE Mesh with the High Speed Video Images and Conduct Digital Image Correlation Analysis
7. Post-processing of Displacement and Strain Data
8. Fine Tuning and Refinement of Results
Before the speckling process, the femur is cleaned from excess fat and soft tissues, and the greater trochanter is potted in an aluminum cup. During solidification of polymethylmethacrylate (PMMA), the bone is wrapped in a saline soaked cloth to avoid tissue dryness. Once PMMA is solidified, the bone is cleaned again right before spraying (Figure 1). Then, the bone surface is sprayed or brushed with a water-based plastic white color. Once dried, the white surface is speckled with black color to have a stochastic pattern of black spots on the white background (Figure 2). Once the bone is placed in the testing fixture, the lights and high-speed video cameras are set, and the optimal contrast of the pattern and the focus of the cameras are checked prior to testing (Figure 3). The DIC method requires a high contrast speckling pattern and sufficient illumination. Otherwise, the results may be affected by several issues such as oversaturation of the surface, poor contrast, and dull images (Figure 4). Uncompressed images from the high speed videos are able to be extracted in multiple temporal sampling regimes and the DIC tracking algorithm can be operated through graphical user interfaces (Figure 5). The outline of the femur sample is used to identify the region of interest for strain field estimation (Figure 6A) and for creation of a finite element mesh for strain calculation (Figure 6B). The onset of fracture is detected by monitoring the degree of strain deviation during testing, with peaks representing bone damage and time frame of fracture (Figure 6C). Finally, 2D strain fields are superimposed back onto the untested bone image for enhanced visualization (Figure 6D).
Figure 1: Bone Preparation before Painting the Bones. (A) cleaning the bone from fat and moisture after being thawed; (B) potting greater trochanter; (C) cleaning before spraying process Please click here to view a larger version of this figure.
Figure 2: Painting Process. (A) DIC working area and necessary tools; (B) spraying bone with white primer; (C) brushing white color on the bone surface; (D) speckling black spots on the white bone surface; (E) final speckled surface of the bone ready for testing Please click here to view a larger version of this figure.
Figure 3: Lighting and Camera Settings. (A) setting up lamps and shields; (B) setting up high speed video cameras; (C) a bone sample loaded to the testing machine with lights and cameras ready for check and test; (D) checking the images for the functionality of the cameras; (E) examining the area of interest, in the femoral neck, for focusing zones, depth of field, lack of blur, and general quality of images for DIC Please click here to view a larger version of this figure.
Figure 4: Bone DIC Caveats. (A) oversaturation on the head region; (B) mixing and flowing of black and white when white surface is not dry; (C) poor contrast, local oversaturation, poor clarity of the image Please click here to view a larger version of this figure.
Figure 5: Custom Scripting Dialogs that were used in the DIC Processing. (A) mov_frames.m, (B) rrImageTrackGui.m, (C) generated 2D mesh Please click here to view a larger version of this figure.
Figure 6: Example of Intermediate DIC Results. (A) spline drawn to highlight the region of interest, (B) generated mesh overlaid on bone image, (C) strain deviations as a function of high speed video frame, (D) calculated strain contour plot associated with the 2 test images prior to bone fracture Please click here to view a larger version of this figure.
We introduced a protocol to consistently prepare femoral samples for high contrast imaging during fracture testing which were then used to estimate full field strain distributions with DIC. This protocol ensured appropriate contrast texture of black tracking speckles against a solid white background on the bone surface. Following this protocol, we successfully replicated the estimation of strains using DIC analysis for eighty-nine femora.
DIC is an optical method which involves placing a mesh over a series of images captured by high speed video cameras and tracking the pixel intensity changes between frames using a cross-correlation algorithm. During the experiments, we found several considerations that need to be taken into account for the accuracy and robustness of the method and those details are reflected in the presented protocol in detail. First, we found the sensitivity and the resolution of cameras are of great importance for the spatial strain measurements of interest. Second, a very fine texture of contrasting black marks on white surface should be avoided as they may not be visible to the cameras. Third, cameras and lighting should be set at appropriate distances to ensure optimal aperture size for depth of field and the quality and contrast of the images. Excessive lighting may lead to saturation of the images resulting in poor contrast. Finally, the temporal spacing between images needs to be set such that the surface speckles do not move more than 6 pixels between frames so that tracking is accurately captured during cross-correlation.
As demonstrated in this work, DIC has the capability to provide full field time-sequence strain estimates for femur fracture tests, something not easily obtained with strain gauge experimental techniques. Although strain gauge measurements have been employed by a number of researchers, such measurements can be hindered by inadequate mounting adhesion to the bone surface, gauge conditioning, and a limited spatial distribution12,13. In contrast, full-field strain data is extremely useful for validation of QCT/FEA models of bone strength by comparing strain fields between model and test, and it also has clinical application to correlate femoral fracture types with the pattern of strain development on the surface of the femur for this physiological fall load case5,9. While fixture compliance could be a problem when testing very stiff femora, DIC circumvents this issue by calculating cortex strains directly from bone local deformation thus, eliminating fixture compliance as a source of errors when estimating femoral stiffness. The results from these image correlations may aid in developing better QCT/FEA models including material failure and metrics of damage and fracture. These can eventually help guide therapy decisions especially for osteoporotic patients.
The method does have several drawbacks, however. The bone specimen surface must be uniformly covered with a stochastic speckle pattern which has high contrast with the background. Occasionally reflections from lighting or large deformations can alter the ability for the algorithm to track the pattern precisely from frame to frame (Figure 4). A second limitation is when single camera (2D) DIC is employed, strain calculations can be affected where the bone surface plane deviates from being parallel with the camera image sensor plane14. This can occur when the femoral surfaces rotate towards or away from the camera during fracture testing. We are exploring future work in this area to add a second camera and utilize 3D DIC methods for improved accuracy. Until recently, such methods have been out of reach in a research setting but are now becoming more readily available. Another limitation of the method specific to biological tissue is the uncertainty of paint adhesion to the femur surface. By our observations, this was not an issue in our testing, but any slippage of the femur tissue and paint would affect the results. Additionally, any non-bone tissue left behind during bone preparation can interfere with cortex strain measurements. Finally, the image tracking settings and mesh density are factors that may affect the quality of results from the DIC analysis and need to be carefully considered.
The current protocol presents a method to efficiently and consistently prepare femoral specimens for digital image correlation analysis and for estimation of corresponding strain fields from high speed camera imaging during fracture testing. It has been demonstrated in our laboratory to yield consistency over multiple testing timeframes and with varying research personnel and operators over a 6 year time period. The procedure for DIC presented here for femoral preparation and testing can be easily extended to other bone types.
The authors have nothing to disclose.
The authors would like to thank the Materials and Structural Testing Core at Mayo Clinic for their technical support in performing the fracture testing. In addition we would like to thank Ramesh Raghupathy and Ian Gerstel for their assistance in developing the DIC scripts and specific details of the DIC protocol during their tenure at Mayo Clinic, and the Victor Barocas Research Group, University of Minnesota for the underlying open source software that performs the core of the digital image correlation strain calculations11. This study was financially supported by the Grainger Innovation Fund from the Grainger Foundation.
Krylon plastic primer white | Krylon, Peoria, AZ, USA | N/A | Used as a base coat for a smooth white finish on bone surface |
Water-based acrylic white and black paint | Plaid Enterprises (Ceramcoat), Norcross, GA, USA | N/A | Paint source for white and black colors |
Mixing bowl | Not specific (generic) | N/A | Used to mix and prepare paint |
Foam brush | Linzer Products, Wyandanch, NY, USA | N/A | Used to apply paint on bone surface |
Toothbrush | Colgate-Palmolive, New York, NY, USA | Firm bristle | Used to apply appropriate size and distribution of speckling pattern |
Hygenic Orthodontic Resin (PMMA) | Patterson Dental, St Paul, MN, USA | H02252 | Controlled substance and can be purchased with proper approval |
Kenmore Freezer | Sears Holdings, Hoffman Estates, IL, USA | N/A | Used to maintain a -20oC storage enviroment for bone specimens |
Physiologic Saline (0.9% Sodium Chloride) | Baxter Healthcare, Deerfield, IL, USA | NDC 0338-0048-04 | Used for keeping specimens hydrated |
Scalpels and scrapers | Aspen Surgical (Bard-Parker), Caledonia, MI, USA | N/A | Used to remove soft tissue from bone specimens |
Fume Hood | Hamilton Laboratory Solutions, Manitowoc, WI, USA | 70532 | Used for ventilation when preparing PMMA for potting of specimens |
Lighting units | ARRI, Munich, Germany | N/A | Needed for illumination of target for image capture |
High-speed video camera | Photron Inc., San Diego, CA, USA | Photron Fastcam APX-RS | Used to capture the high speed video recordings of the fracture events |
Photron FASTCAM Imager and Viewer | Photron Inc., San Diego, CA, USA | Ver.3392(x64) | Used to record and view the high speed video recordings |
Camera lens | Zeiss, Oberkochen, Germany | Zeiss Planar L4/50 ZF Lens | Needed for appropriate image resolution |
ABAQUS CAE | Dassault Systemès, Waltham, MA, USA | Versions 6.13-4 | Used for defining region of interest and creating finite element mesh |
MATLAB | Mathworks, Natick, MA, USA | Version 2015b | Used for image processing and DIC analysis |
TecPlot | TecPlot Inc., Bellevue, WA | Used for post processing of strain fields | |
Strain Calculator Software | Victor Barocas Research Group, University of Minnesota, Minneapolis, MN, USA | http://license.umn.edu/technologies/20130022_robust-image-correlation-based-strain-calculator-for-tissue-systems | Used to calculate strain field |
mov_frames.m | Matlab script, Mayo Clinic, Rochester, MN,USA | N/A | Used to downsample uncompressed images from high speed video files |
convert_imagesize.m | Matlab script, Mayo Clinic, Rochester, MN,USA | N/A | Used to register image pixel coordinates with mesh coordinates |
rrImageTrackGui.m | Matlab script, Mayo Clinic, Rochester, MN,USA | N/A | Used to perform the image cross-correlation to obtain deformations and run Strain Calculator |
analyzeFailurePrecursor.m | Matlab script, Mayo Clinic, Rochester, MN,USA | N/A | Used to track the peak strain components temporally |
makeMovies.m | Matlab script, Mayo Clinic, Rochester, MN,USA | N/A | Used to create portable *.avi movies of the deformation components, strain components, principal strains, von Mises strain, and strain energy |