JuliaHealthZoo

Appearance

Interoperability (PythonCall/C++) ​

In MRI research, interoperability is not just a convenience layer. It is often the practical bridge between simulation, reconstruction, visualization, and existing lab tooling. Teams rarely start from a blank slate. Some components already live in Julia, some in Python, and many mature reconstruction tools are still C/C++ based.

For this reason, this workflow treats Julia as the orchestration language and uses interoperability deliberately: keep data movement explicit, keep dependencies project-scoped, and keep the reconstruction path reproducible across machines.

Core interoperability principles used here:

  • keep language boundaries explicit

  • keep environment setup project-local

  • keep data exchange formats visible and reproducible

PythonCall provides the Julia/Python bridge used here. It supports importing Python modules directly from Julia, calling Python functions, and converting values explicitly with pyconvert when Julia-native types are required.

It also provides wrapper types around Python objects and integrates with CondaPkg for project-level dependency management.

Minimal PythonCall pattern ​

julia
using PythonCall

np = pyimport("numpy")
arr = np.arange(8)

If a BART Python module is available in the environment, it can be imported in the same style and called from Julia.

julia
using PythonCall

bart_py = pyimport("bart")
out = bart_py.bart(1, "version")

This directly demonstrates the "import and call BART Python functions from Julia" path.

C/C++ toolchain interoperability context ​

In this workflow, BART represents the C/C++ side of the ecosystem.

In practice, Julia can interact with BART through three main paths:

  • BartIO.jl for Julia-native command calls and .cfl/.hdr handling

  • direct CLI calls for command-level reproducibility

  • Python wrapper access through PythonCall when Python helpers already exist

When to use what ​

  • Prefer BartIO.jl when:
- you want Julia-first scripts
- you need direct `.cfl/.hdr` handling
- you want low cross-language overhead
  • Prefer PythonCall when:
- your team already uses Python notebooks or helper modules
- a required component is available only in Python
- hybrid prototyping speed is the priority
  • Use direct CLI calls when:
- validating exact BART command behavior
- matching published command pipelines
- debugging wrapper-independent differences

Official sources ​