From Calf to 3D Skeletal Model: The Journey of a Moose Calf (Alces alces)
Project summary
There appeared to be a gap in knowledge regarding the preparation and analysis of, and access to, neonatal mammalian skeletons. Such skeletons do not appear to be commonly curated, especially not in intact/untarnished form. When a neonatal moose calf came into the possession of Dr. Roy Rea at UNBC, he saw an opportunity to curate a unique specimen and provide a rare learning opportunity.
The first challenge was regarding how to properly prepare and re-articulate a neonatal skeleton, and second, provide wildlife biologists and the wildlife medical community procedures on how to accomplish this. This project aimed to help fill this information gap and provide a valuable resource to relevant professionals.
To complete this project, a number of individuals were involved. Initially, student Lena Richter (under the supervision of Dr. Rea) and a local taxidermist extracted the skeleton from the calf carcass. The cape (skin) was mounted and taxidermied by Rod Gray (Bear Bone-z Taxidermy) while Lena got to work on cleaning the bones. After the bones were cleaned, Holly McVea (under the supervision of Dr. Rea) prepared and articulated the skeleton into a viewable educational mounted display with guidance and assistance from Bill Jex (Biologist), Lee Post ("The Boneman"), Catherine Pelletier (Anthropology student), and Doug Thompson and John Owlowski (Enhanced Forestry Lab Greenhouse curators). Last, Rod McLatchy led the effort to create the 3D model using photogrammetry techniques.
The resulting documented process, mounts, and 3D models have been designed to act as tools in the fields of wildlife biology and wildlife medicine in both practice and education in a free and accessible manner.
To download the 3D model of the full moose neonate, click here.
To download the 3D model of the moose neonate's skull, click here.
If you would like more detailed instructions, steps, supplies, and sketches, see Holly's notebook.
For any other information or photos, please email roy.rea@unbc.ca.
Figure 1. Images of the taxidermied neonatal moose and its articulated skeleton.
Introduction to the calf
A one- to two-week-old male moose calf was found near Prince George, British Columbia, Canada tangled in a barbed wire fence by a group of hikers on June 6, 2017. At the time of discovery, the calf was alive but all alone. The hikers freed the calf and took it to the Ospika Veterinary Clinic, in nearby Prince George. The skin and connective tissues near the bottom of the calf’s right rear leg had been badly damaged. The vets at the vet clinic worked hard to bandage it up and get the calf rehydrated. When the calf was stabilized, co-author Roy Rea transported it to a rendezvous point with staff from the Northern Lights Wildlife Shelter in Smithers, B.C. There, the calf was given antibiotics and cared for and lived for a few days after which it eventually succumbed to its injuries on June 9th. The shelter believed the calf eventually died of capture myopathy related to its struggle in the fence. Following discussions with the shelter, we had the calf frozen whole. Several days after its death and freezing, we transported the calf back to the university where it was stored in a -20 °C freezer until thawing and processing.
Figure 2. The moose calf after he had been brought to Northern Lights Wildlife Shelter in Smithers, B.C.
Bone cleaning
The calf was fully thawed and skinned by a local taxidermist (Rod Gray - Bear Bone-z Taxidermy), who volunteered his time and expertise to create a full taxidermied hide mount for display alongside the re-articulated skeleton. The skinned carcass was then taken to the UNBC lab and was disarticulated into sections (legs, skull, spinal column with ribs) using necropsy knives. The bulk of the flesh was removed using knives and scalpels. Sections that were not being dissected were re-frozen until needed. Long bones of the limbs including scapulae and all associated joint structures were cleaned by trialing two different methods. The first was using a dilute hydrogen peroxide solution (1-2%). Bones were soaked for 3 hours, then removed and the softened flesh was scraped off. This process was repeated until all flesh was removed. This method was chosen as it is gentle on the immature bones and minimizes damage to epiphyseal plates. It is, however, tedious in nature. Therefore, a second method (boiling) was used for the remaining leg bones. Bones were gently boiled in a large stock pot until protein was denatured and flesh could be readily detached from the bone. This method proved to be much more efficient, but it did cause mild damage to the less dense bone ends. Photographs were taken throughout this process to document the organization of the bones before they were disarticulated. Bones were kept sorted to their respective “limbs” by placing the loose parts in labeled mesh laundry bags during boiling/soaking.
Due to their more fragile and intricate nature, the skull, spinal column, ribs, and pelvis (still connected by cartilage and connective tissue) were placed in a dermestid beetle tub for cleaning (kindly done by Kelly - Dark Arts Oddities). This process can take several weeks depending on how much soft tissue is on the bones. Once cleaned, these bones were also placed in a 1-2 % hydrogen peroxide soak for 2-3 hours to remove staining. Great care was taken to handle these bones only in a separate building from any taxidermied lab specimens to prevent accidental dermestid beetle infestation.
Post cleaning, all bones were soaked in a 2% sodium borate (20 Mule Team Borax) solution (as much borax to water as will dissolve) to degrease.
Figure 3. Moose calf bones soaking in a 1-2 % hydrogen peroxide to remove staining.
Bone preparation
Initial preparation of the bones began with an examination of the condition of the bones, looking at the amount of left-over cartilage, and creating a plan to prepare the bones for permanent mounting/longevity without a significant amount of deterioration. The larger bones, such as the leg bones, had their epiphyseal caps removed and were drilled through on both sides to remove bone marrow; bone marrow can leak out and cause staining over time. Once most of the bone marrow had been scraped out of the bones, the larger bones and caps (as well as the vertical column, as it had a large amount of connective tissue still attached) were left to soak in a solution of diluted ammonia and dish soap for a couple of weeks (1 gal of 5% ammonia and ½ cup dish detergent into 4 gal of water). The bones were deemed ‘sufficiently soaked’ once the connective tissue had softened (and was easy to remove) and the water was clear of floating fat and residue; the solution had to be drained once and refreshed with more water and dish soap due to the amount of lipids removed. After the bones had completed their ammonia soak, they were cleaned by scalpel and scrub brush. After cleaning, the bones were oven-dried at temperatures between 35 - 65 °C over 3 days.
Once dry, many of the soft calf bones were fragile (given their lack of ossification) and decaying. In an effort to preserve and harden the bones, they were dipped 2-3 times (15 s each time) in a bone-hardening solution consisting of school glue and water; the bones were hung using fishing hooks and line or placed on plastic or paper sheets to dry either on the counter or in a fume-hood to allow for more air current to expedite drying. Some bones with finer or more delicate features were at risk of the glue pooling in undesirable places and so were 'painted' with glue, or dipped and then were ‘painted’ to remove excess glue. The bones, once dry, were sanded and cleaned of excess glue drops and then re-organized into sections in preparation for articulation.
Bone articulation
The first section to be put together was the verebral column/spine. A threaded rod (0.5 in) was bent to the approximate shape of the spine using a rail track, a mallet, a metal post, a table, and a bench vice; The Moose Manual by Lee Post (The Boneman) states that the rod should be 1/3 the diameter of the smallest vertebra. Approximations were derived from diagrams in The Moose Manual and from measurements taken from the taxidermied calf; see a sketch with measurements of the calf in the “Notes” section of Holly’s notebook. The lumbar, thoracic, and the cervical vertebrae (aside from the atlas and the axis) were sorted, glued together (as many were broken into 2-3 pieces), marked with a pen, and drilled through (from both ends, meeting in the middle as to lessen the stress on the bones and avoid breakages) and threaded onto the spinal rod as you would with beads; the vertebrae were secured in place using hot glue with spacing in between the bones to account for connective tissue. Before the 4th cervical vertebra, a piece of metal electrical conduit flattened with a mallet and drilled through was threaded onto the spinal rod to act as a stand mounting attachment point.
Once the spine was assembled, the pelvis and tail bones were assembled. The pelvis was put together using a combination of hot glue and epoxy. Once dried, the ilia were drilled and bolted to a flattened end of a metal electrical conduit post with a hole drilled through it; final positions were secured using hot glue. Once attached to the post, the pelvic section was attached to the spine. The complete spine and pelvis were then mounted onto the stand; the stand consisted of a wooden base with all-threaded rod posts bolted to it. The stands had washers and nuts on the posts to allow for height adjustment and in an effort to avoid welding.
Once the spine and pelvis were mounted, rib cage assembly began. The six-piece unfused sternum was configured using epoxy and wire. Once dry, the sternum was drilled to allow for wire to connect the ribs to the sternum. While that dried, the ribs were connected. All of the ribs were missing caps (and many had broken ends); ends and caps were refashioned (with assistance from Catherine Pelletier) to the bone using apoxie sculpt smoothened with bubble blowing solution to lubricate sculpting utensils and fingers (recommended by Lee Post). After the bones dried, they were drilled on both ends and hot-glued to the sternum and vertebrae (these attachments were reinforced with more wire and silicone). The sternum was not yet completely ossified. Therefore some of the cartilaginous sections were removed during cleaning, and the connective tissue was removed; wire and silicone were then used to recreate the length of the missing section needed for all ribs to attach as they once were to a central body.
Once the rib cage was complete, the legs were built. The bones were organized by leg, and then ordered for drilling and connecting. The bones were connected using two lengths of twisted wire per bone intersection. Many of the bones were damaged (due to drilling, removal of caps for cleaning, etc. given the crumbly nature of the bones even with the assistance of the bone hardening solution) so more apoxie sculpt was used to hide any discolouration, conceal dried tissue that would not come off, and correct breaks or missing bones and cover glue. Knee rearticulations were a challenge, as they were composed of 5 small undefined and unossified bones that were difficult to orient. Once positioned, they were glued, drilled, and mounted into place; recreation was not attempted for the cartilage in the knees as it would obstruct the nature of the bones present. Legs were either mounted using 1 nail and hot glue (front legs) or bolted and glued (hind legs).
To finalize the articulation, the skull was mounted to the front-most stand by putting the post through its foramen magnum into the intracranial space that was filled with hot glue. The lower jaw was attached to the skull with finishing nails through the mandible and a final finishing nail was placed between the upper and lower jaw to show the inside of the mouth/teeth. After the skull was placed, apoxie sculpt was used to correct any missing/broken pieces, and acrylic paint was used to conceal the breaks and cover the apoxie sculpt. Between two and four coats of paint were used on areas of discoloration; some paint was also scrubbed away to help give a more patchy-natural look to the skeleton. Once dry, the skeleton was positioned using hot glue, wire, and one more stand mount (placed under the left front limb) to mimic the position of the taxidermied calf as a direct comparison for bone positioning to the taxidermied positioning as a way to provide a useful education tool. After the mount was finalized, a sign denoting the students who worked on the project, as well as the species and approximate age of the specimen, was mounted to the stand using slanted wood and glue.
Figure 4. Moose calf vertebrae pre-drilled to be mounted onto the spinal threaded rod.
Skeletal photogrammetry: 3D model creation
Photogrammetry mimics the way we see 3D with stereoscopic vision. Our eyes independently see a slightly different image; this enables us to estimate what object/feature is closer and which is further. This is also referred to as parallax. Photogrammetry software builds 3D models by measuring details found in comparing multiple photographs. Each photo needs to be taken at a slightly different position, to compare if a detail is closer or further. Each photo needs to be taken in small, short offsets from each other. If photographs are taken too far apart, found features will be too different for the software to declare them as the same feature in each photograph. To create a good set of photos for a scan, we need to ensure we have enough photos to capture all sides of an object, and enough photos to continually match features to relate the photographs to each other. Accurate consistent photographs are essential when doing photogrammetry. Photographs with blur (or noise) introduce distortions or, worse, do not allow computers to find features to measure and compare with. Inconsistent lighting and reflections can also prevent feature matching. To acquire the most detail possible in the model, photographs with a high resolution and a large depth of field are ideal. A large depth of field can be achieved by using a narrow aperture. To optimize the photographs, we used a Nikon Z5 mirrorless camera with a 24-70 mm lens set at 24 mm F13 1/3s ISO 200 mounted on a tripod.
Reality Capture is the software we used for our final model. After importing the photographs, the software estimates the position (and orientation) in which each photograph was taken, using details found in the photos and details listed about the camera used to take the photo. The software then finds all the features and measures their position and colour. These features are referred to as a point cloud since the measured points coalesce along the surface of the object. Very commonly, the whole room will be represented in the point cloud. Features will be detected anywhere in the photos, including objects detected in the background.
After the point cloud is generated, the 3D model can be created. These 3D models are a collection of triangles that define the surface of our scan. In the case of our moose skeleton, our point cloud contained ~6 million points and the model consisted of ~15.1 million triangles. Reality Capture has some tools to reduce model size by combining triangles. For extra detail, a texture image can be added to the triangles. This texture image draws an image to display on each triangle, instead of displaying a single solid colour.
Since this dataset is millions of points in 3D, modern gaming computer’s video cards are perfectly suited to deal with this type of processing. Nvidia's Cuda-based video cards have thousands of parallel processors; this is perfect for computing this large quantity of complex data much more quickly as opposed to a CPU handling these calculations one point at a time. Depending on the scan complexity and number of photographs, this processing can still take hours!
Initially, when we started the scanning project, we used a GTX 960 4 GB video card. This card has 1024 parallel processors. For the final scan, we processed it with an RTX 3070 8 GB video card (with 5888 parallel processors). The latter of the cards cut processing time by more than 10 x, which allowed us to make adjustments and add photos more efficiently to create the final model. However, please note that just as there were upgrades in hardware, the newer version of software has provided better models faster.
Figure 5. A gradation from individual 3D points (the pointcloud) to the complete 3D model of a neonatal moose calf skeleton.
Acknowledgements
We would like to thank the following businesses and people for their donations of time, tools, products, and any other contributions to the project:
Ospika Veterinary Clinic, Rod Gray (Bear Bone-z Taxidermy), Ken Otter (UNBC), Kelly (Dark Arts Oddities), Northern Lights Wildlife Shelter, The Mabbett Family, The Adams Family, Dakota Den Duyf, The UNBC Student Life Department, Doug Thompson and John Orlowski at the Enhanced Forestry Lab (EFL) at UNBC, Northern Hart Design (Thomas Torraville), Lee Post (“The Boneman”), Catherine Pelletier, Farid Rahemtulla, Michelle McLatchy, Kyle Ross, Dan Aitken, and Bill Jex.