Pulse-sequence and protocol management¶

Sequences in Adelpha MRI are MRI4ALL plugins, not a new pulse-sequence language. The Imaging Console lists them, edits parameters, and writes the exam queue. Python SequenceBase still builds the waveform.

Adding inputs from Python¶

The Imaging Console SEQUENCE / ADJUSTMENTS / SYSTEM / PROCESSING / OTHER tabs are filled from Python. You do not write TypeScript or a Qt .ui file.

Name every operator field param_<Name>. param_TE becomes the TE box. Known names (TE, TR, FOV, …) pick up units from a shared table. Anything else still appears as a number, checkbox, or text field.

from sequences import PulseqSequence, param
from pathlib import Path

class MySequence(PulseqSequence, registry_key=Path(__file__).stem):
    param_TE: int = 10          # SEQUENCE tab, unit ms (known name)
    param_TR: int = 250
    param_coarse_steps = param(
        100,
        title="Coarse steps",
        tab="adjustments",      # or sequence / system / processing / other
        minimum=1,
        description="Points in the first search",
    )
    param_mode = param("fast", enum=["fast", "slow"], tab="other")

param() sets label, unit, tab, min/max, dropdown enum, and optional help text. After you save the file, restart the Python runtime so the sequence catalog reloads.

If you do not implement get_default_parameters / get_parameters / set_parameters, Adelpha copies the param_* values for you. Existing sequences that still map those methods by hand keep working.

Optional overlay without converting an attribute to param():

parameter_ui = {"N_ITER": {"title": "Iterations", "tab": "adjustments", "minimum": 1}}

See sidecar/console/sequences/adj_frequency_snr.py for a sequence that exposes search controls on the ADJUSTMENTS tab this way.

Catalog the GUI sees¶

sidecar/console/services/api/sequences_api.py loads SequenceBase when the MRI4ALL environment is complete. If numba / numpy pins block that import (the desktop sidecar uses numpy 2), the façade serves a FALLBACK catalog so the UI still lists sequences:

Plugin sources live in sidecar/console/sequences/ (for example se_2D.py). Parameter schema comes from common.parameter_schema.schema_for_defaults. Validate with the façade’s validate route before Prepare.

Exam queue¶

flowchart TD
  A[Patient Registration] --> B[Open exam in memory]
  B --> C[Sequence list: add / rename / duplicate / delete]
  C --> D[Parameter tabs: SEQUENCE ADJUST SYSTEM PROCESSING OTHER]
  D --> E[Prepare writes scan.json into PREPARED]
  E --> F[In-process pipeline runs SequenceBase]

Restarting the MRI API clears the in-memory exam. Protocols in the hamburger menu (Exam → protocols) are MRI4ALL protocol browser flows, rehosted in the React shell. Right-click on the sequence list to rename, duplicate, or delete entries for the open exam.

What stays on disk¶

The TypeScript client talks only to /api/mri (desktop: supervisor gateway + session token). It never opens the Red Pitaya socket. Operator steps: Imaging Console. Hardware: MaRCoS.