dmi_earl.py

# small demo that shows how to fit radial profiles of NEMA sphere phantom to estimate
# resolution of a PET reconstruction

# we do the fit 3 times with different settings (fixing a different amount of parameters)
# the 3rd approach should be the most accurate if the sphere radii are known

import os
import pymirc.fileio as pmf
import pymirc.viewer as pv

import pynemaiqpet
import pynemaiqpet.nema_wb as nema

from scipy.ndimage import gaussian_filter
from glob import glob

from argparse import ArgumentParser


parser = ArgumentParser(description="Analyze GE PET NEMA in terms of recovery")
parser.add_argument("dcm_dir_pattern", help="dicom direcotry pattern")
parser.add_argument("--dcm_file_pattern", default="*.dcm", help="dicom file pattern")
parser.add_argument(
    "--injected_signal",
    default=None,
    type=float,
    help="injected signal in the spheres. if not given, taken from biggest sph.",
)
parser.add_argument(
    "--sm_fwhm_mm",
    default=5,
    type=float,
    help="FWHM of Gaussian kernel (mm) used for additional post-smoothing",
)
parser.add_argument(
    "--force_smoothing",
    action="store_true",
    help="force smoothing. even for already smoothed data",
)
parser.add_argument(
    "--earl_version",
    default=2,
    type=int,
    choices=[1, 2],
    help="EARL version to use for limits in recovery plots",
)
args = parser.parse_args()


# injected sphere activity concentration at start of acq.
injected_signal = args.injected_signal
sm_fwhm_mm = args.sm_fwhm_mm
force_smoothing = args.force_smoothing
earlversion = args.earl_version

dcm_dir_pattern = args.dcm_dir_pattern
dcm_file_pattern = args.dcm_file_pattern

# ------------------------------------------------------------------------------------------------

dcm_dirs = glob(dcm_dir_pattern)


for dcm_dir in dcm_dirs:

    print(os.path.basename(dcm_dir))

    # load example data set included in package
    dcm_pattern = os.path.join(dcm_dir, dcm_file_pattern)
    dcm = pmf.DicomVolume(dcm_pattern)
    vol = dcm.get_data()
    voxsize = dcm.voxsize

    # FWHM of Gaussian kernel to apply before analysis
    dcm_hdr = dcm.firstdcmheader

    # check if the data was already post-smoothed by analysing private GE tags
    # for trans-axial and axial filter
    if (
        (dcm_hdr[0x0009, 0x10BA].value == 0) and (dcm_hdr[0x0009, 0x10DB].value == 0)
    ) or force_smoothing:
        sm_str = f"_{sm_fwhm_mm}mm_ps"
    else:
        sm_fwhm_mm = 0
        sm_str = ""

    # post smooth the image
    if sm_fwhm_mm > 0:
        print(
            f"Applying additional isotropic Gaussian post-smoothing of {sm_fwhm_mm}mm"
        )
        vol = gaussian_filter(vol, sigma=sm_fwhm_mm / (2.35 * voxsize))

    # -------------------------------------------------------------------------------------------------
    # do a fit where we force all spheres to have the same signal (assuming that the activity
    # concentration in all sphere was the same
    # and we fix the radii of all spheres (using the values from the NEMA specs)

    # try also doing the fit without fixing the radii
    fitres, sphere_results = nema.fit_WB_NEMA_sphere_profiles(
        vol,
        voxsize,
        sameSignal=True,
        Rfix=[18.5, 14.0, 11.0, 8.5, 6.5, 5.0],
        showBGROI=False,
        Sfix=injected_signal,
    )

    print("fit with same signal and fixed radii")
    print(sphere_results)
    sphere_results.to_csv(
        os.path.join(
            dcm_dir,
            f"{dcm.firstdcmheader.SeriesDescription}{sm_str}.csv".replace(" ", "_"),
        )
    )

    # -------------------------------------------------------------------------------------------------
    # show the results

    fig = nema.show_WB_NEMA_profiles(fitres)
    fig.savefig(
        os.path.join(
            dcm_dir,
            f"{dcm.firstdcmheader.SeriesDescription}{sm_str}_profiles_EARL_{earlversion}.png".replace(
                " ", "_"
            ),
        )
    )

    # plot the max and a50 recoveries
    # the 2nd argument should be the true (expected) activity concentration in the spheres
    # and also show the limits given by EARL (vesion 2)

    if injected_signal is None:
        ref_signal = sphere_results.signal.values[0]
    else:
        ref_signal = injected_signal

    fig2 = nema.show_WB_NEMA_recoveries(
        sphere_results, ref_signal, earlversion=earlversion
    )
    fig2.suptitle(f"{dcm.firstdcmheader.SeriesDescription}{sm_str}")
    fig2.tight_layout(pad=2)

    fig2.savefig(
        os.path.join(
            dcm_dir,
            f"{dcm.firstdcmheader.SeriesDescription}{sm_str}_RCs_EARL_{earlversion}.png".replace(
                " ", "_"
            ),
        )
    )