Great news for all users of ImplaStation software!
We recently implemented a significant update, namely the addition of segmentation from DICOM data.
Let’s ask you: how often did you get an accurate bone surface after converting the results of CT by optical density into a 3D model? In our practice, we often have inaccuracies, holes, and artifacts. Are these results enough? Definitely not. The model is more suitable for demonstration, and if you make, for example, guides or membranes with support on the bone, then additional processing is still needed.
Segmentation tools help achieve the desired accuracy
Segmentation tools help achieve the desired accuracy.
Pic.1
If you have extra time and patience, you can use manual segmentation, but if every minute is valuable, then you should pay attention to the possibilities of artificial intelligence. It significantly reduces the processing time of materials, and as a result, the model turns out to be more like a real bone. But still, the obtained result is not ideal either. At first glance, the surface is excellent, without holes and artifacts, but in the section, inaccuracies become noticeable. Why so? Because the neural networks do not yet fully take into account nuances and details that may become critical in further work.
Segmentation is now also possible in the ImplaStation software
We have installed a tool that edits the bone in semi-manual/semi-automatic modes.
Pic.2
This, of course, may take some time, but the resulting surface is the most accurate.
Pic.3. Comparison of results of conversion and segmentation by AI tools
However, the speed can be increased by combining the software tool with the outcome of neural network processing.
Pic.4 Comparison of AI and ImplaStation segmentation results
By the way, we are currently actively working on our own artificial intelligence library for the even greater functionality of ImplaStation.
So, pay attention to our update and expect a tutorial soon with a description of all the features of the new feature.
Best regards,
the ImplaStation team
