1- There is no automatic and built-in way to do spine segmentation in MR images.
Since there is no single workflow that would fit any kind of segmentation in any modality, we recommend to try the different tools and tune their parameters. For instance you can adjust the size, the adaptiveness as well as the smoothing of the brush. Also keep in mind that the brush takes into account the current window/level (contrast/brightness) of the image, so you can also try to adjust that to maximize the visibility of your target structure.
On top of that, the next version of Labels will include a real-time rendering of the brush stencil, which makes the process nicer and easier to control.
You can also speed up the annotation process if you already have an algorithm in Python (for instance a machine learning model) that you can run via the Run Script action - see documentation shipped with Labels.
2- I am not sure to understand what you are asking: do you want to use ImFusion algorithms in your code, or the other way around? Can you give us more details on what you want to do?