Cardiac CT Phases

A cardiac CT folder often holds many phases of the heart cycle in one DICOM series. This tutorial finds the phases, loads one, and runs the pipeline on each phase.

1. Find the Phases

from pictologics.utilities import get_dicom_phases

phases = get_dicom_phases("cardiac_ct/")
for phase in phases:
    print(phase.index, phase.label, phase.num_slices)
# 0 Phase 0% 160
# 1 Phase 10% 160
# ...

get_dicom_phases splits the series by the first of these tags that changes: the cardiac phase percentage, the temporal position, the trigger time, the acquisition number, the echo number; else by repeated slice positions. A tag splits the series only at repeated slice positions, so a scanner that writes a new acquisition number every few slices gives one phase.

2. Load One Phase

from pictologics import load_image

image = load_image("cardiac_ct/", dataset_index=4)  # the 5th phase
image = load_image(phases[4])                        # the same, without reading the folder again

A folder with more than one image series (for example two reconstructions) needs series_uid; the error lists the series.

3. Run Each Phase

For an image path, image_options gives options to load_image. So the pipeline loads each phase from the folder path:

from pictologics import RadiomicsPipeline, format_results, save_results

pipeline = RadiomicsPipeline.from_template("lv")
rows = []
for phase in phases:
    results = pipeline.run(
        "cardiac_ct/",
        f"masks/lv_phase{phase.index}.nii.gz",
        image_options={"dataset_index": phase.index},
        config_names=["lv_fbn_32"],
    )
    rows.append(format_results(results, meta={"phase": phase.label}))
save_results(rows, "lv_by_phase.csv")

The log entry of each configuration records the image_options. In run_batch, give each phase as its own case with its image_options (see Many ROIs and Batch Studies).

4. The Cardiac Templates

Two templates hold configurations for cardiac CT:

pipeline = RadiomicsPipeline.from_template("coronary")
print(pipeline.list_configs())

Look at the steps of a configuration with pipeline.get_config(name), and change them for your study.

See also: Data Loading and Rings Around an ROI (the fat around a vessel).