To begin, use an osteochondral autograph transfer system to create a 4.75 millimeter diameter defect in the lateral condyle of a euthanized sheep. Using an orthopedic hammer, create a defect 10 millimeters in depth, and retrieve the osteochondral explant from the condyle. Prepare the calcium phosphate cement, supplemented with 40 micrograms per milliliter of BMP 2, and load the cement paste into a three milliliter syringe.
Then inject the cement into the explant defect with an 18 gauge needle. Turn on the micro tomography x-ray machine, and place the explant tube on the sample holder. Set the resolution to 10.7 micrometers, and exposure time to 1, 200 milliseconds, with a one millimeter aluminum filter at 80 kilovolts and 125 microamperes.
Average three images for each 0.45 degree rotation increment to enhance the signal to noise ratio. To conduct image segmentation, use the integrated segmentation wizard to train a deep learning model for distinguishing between bone and cement. Select a representative zone containing bone, cement, and background from the reconstructed micro-computed tomography images, and segment this first frame.
Now, in the model tab, generate a deep learning model and select the 3D net routine. Then right click on the generated model, and set the experimental parameters as the depth of five, patch size 32 by 32, algorithm, stride ratio of 0.25. And 10 x data.
Use the segmented frame to train the deep learning model by clicking the train button. Once the training is complete, define a second frame and automatically segment it using the predict function. Then click on export to publish the trained model, and apply it to the entire micro-computed tomography dataset by selecting segment, exported model, segment, full dataset.
