Adopt
Primary language for signal processing and ML inference. Current production stack. Python's dominance in the ML ecosystem (NumPy, SciPy, PyTorch, scikit-learn) makes it the natural choice for an rPPG signal processing pipeline. The data scientist-led development history reinforces this as the correct foundation.
IEC 62304 §5.4 software detailed design requirements apply to the Python ML inference code — it must be documented and tested to a standard appropriate for a Class IIa device.