def test_split_epochs_rt_overlap():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], order='rt')
imgs = img.split_epochs(4, overlap=2)
assert len(imgs) == 3
for idx, img in enumerate(imgs):
if idx in (0, 2):
assert img.ntis == 1
assert img.tis == [1 + idx / 2]
assert not img.have_plds
assert img.rpts == [1]
assert img.ntc == 1
assert img.order == "rt"
data = img.nibImage.get_data()
# Epoch 1 is TIs 1111, data 0123, mean across repeats 1.5
# Epoch 3 is TIs 2222, data 4567, mean across repeats 5.5
start = idx * 2
for z in range(data.shape[3]):
assert np.all(data[..., z] == start + z + 1.5)
else:
assert img.ntis == 2
assert img.tis == [1, 2]
assert not img.have_plds
assert img.rpts == [1, 1]
assert img.ntc == 1
assert img.order == "rt"
data = img.nibImage.get_data()
# Epoch 2 is TIs 1122, data 2345, mean across repeats 2.5, 4.5
for z in range(data.shape[3]):
assert np.all(data[..., z] == 2 * z + 2.5)
def qpdata_to_aslimage(qpd, options=None, metadata=None, grid=None):
"""
Convert QpData to oxasl.AslImage using stored metadata where available
"""
# If metadata is not provided, get the existing metadata
if metadata is None:
metadata = qpd.metadata.get("AslData", {})
# If options are provided, use them to override existing metadata
if options:
for opt in METADATA_ATTRS:
val = options.pop(opt, None)
if val is not None:
metadata[opt] = val
else:
metadata.pop(opt, None)
# Create AslImage object, this will fail if metadat is insufficient or inconsistent
from oxasl import AslImage
if grid is None:
aslimage = AslImage(qpd.raw(),
name=qpd.name,
xform=qpd.grid.affine,
**metadata)
else:
aslimage = AslImage(qpd.resample(grid),
name=qpd.name,
xform=grid.affine,
**metadata)
return aslimage, metadata
def test_multite_mean_across_repeats_ibf_rpt():
d = np.zeros([5, 5, 5, 16])
for z in range(16):
d[..., z] = z
img = AslImage(name="d",
image=d,
plds=[1, 2],
iaf="tc",
ibf="rpt",
tes=[8, 9],
casl=True,
bolus=[3, 4])
imgmar = img.mean_across_repeats(diff=False)
assert imgmar.ntes == 2
assert imgmar.tes == [8, 9]
assert imgmar.ntis == 2
assert imgmar.have_plds
assert imgmar.plds == [1, 2]
assert imgmar.tis == [4, 6]
assert imgmar.rpts == [1, 1]
assert imgmar.ntc == 2
assert imgmar.order == "eltr"
assert imgmar.nvols == 8
for z in range(8):
# e.g. mean of 0th and 8th image = 4
assert np.all(imgmar.data[..., z] == (z + 4))
def test_derived():
d1 = np.random.rand(5, 5, 5, 8)
d2 = np.random.rand(5, 5, 5, 8)
img = AslImage(name="d1",
image=d1,
plds=[1, 2],
iaf="ct",
ibf="tis",
casl=True,
bolus=[3, 4])
assert img.ntis == 2
assert img.have_plds
assert img.plds == [1, 2]
assert img.tis == [4, 6]
assert img.rpts == [2, 2]
assert img.ntc == 2
assert img.order == "lrt"
img2 = img.derived(d2, name="d2")
assert img2.ntis == 2
assert img2.have_plds
assert img2.plds == [1, 2]
assert img2.tis == [4, 6]
assert img2.rpts == [2, 2]
assert img2.ntc == 2
assert img2.order == "lrt"
def test_perf_weighted_tr():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], order='tr')
pwi = img.perf_weighted()
assert (isinstance(pwi, Image))
data = pwi.data
assert list(data.shape) == [5, 5, 5]
assert np.all(np.mean(d, -1) == data)
def test_diff_ct():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1], iaf="ct", order='lrt')
img = img.diff()
assert img.ntis == 1
assert img.tis == [1]
assert not img.have_plds
assert img.rpts == [4]
assert img.ntc == 1
assert img.order == "rt"
data = img.nibImage.get_data()
assert list(data.shape) == [5, 5, 5, 4]
assert np.all(data == -1)
def test_t1im():
"""
Check T1 image priors are correctly handled
"""
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], order="prt")
wsp = Workspace(infertiss=True, inferbat=True, infert1=True)
wsp.t1im = Image(np.random.rand(5, 5, 5))
steps = basil.basil_steps(wsp, img)
assert (len(steps) == 2)
options = _get_defaults(img)
_check_step(steps[0],
desc_text="tissue",
options=dict(options, **{
"inct1": True,
}))
_check_step(
steps[1],
desc_text="T1",
options=dict(
options,
**{
"inct1": True,
"infert1": True,
"PSP_byname1": "T_1",
"PSP_byname1_type": "I",
#"PSP_byname1_image" : "t1file",
}))
def test_onestep():
"""
Check that single step mode works when you would normally get multiple steps
"""
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], order="prt")
wsp = Workspace(infertiss=True,
inferbat=True,
infertau=True,
inferart=True,
spatial=True,
onestep=True)
steps = basil.basil_steps(wsp, img)
assert (len(steps) == 1)
options = _get_defaults(img)
options.update({
"method": "spatialvb",
"param-spatial-priors": "N+",
"PSP_byname1": "ftiss",
"PSP_byname1_type": "M",
"inctau": True,
"incart": True,
"inferart": True,
"infertau": True,
"convergence": "maxits",
})
options.pop("max-trials")
_check_step(steps[0], desc_text="spatial", options=options)
def test_mean_across_repeats_tr():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], order='tr')
img = img.mean_across_repeats()
assert img.ntis == 2
assert img.tis == [1, 2]
assert not img.have_plds
assert img.rpts == [1, 1]
assert img.ntc == 1
assert img.order == "tr"
data = img.nibImage.get_data()
assert list(data.shape) == [5, 5, 5, 2]
for znew, zold in enumerate([3, 4]):
assert np.all(data[..., znew] == zold)
def test_reorder_lrt_tlr():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], iaf="tc", order='lrt')
img = img.reorder("tlr")
assert img.ntis == 2
assert img.tis == [1, 2]
assert not img.have_plds
assert img.rpts == [2, 2]
assert img.ntc == 2
assert img.order == "tlr"
data = img.nibImage.get_data()
assert list(data.shape) == [5, 5, 5, 8]
for znew, zold in enumerate([0, 4, 1, 5, 2, 6, 3, 7]):
assert np.all(data[..., znew] == zold)
def test_pvc():
"""
FIXME we need to test the PVC initialization step
and how to do this is not finalized
"""
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], order="prt")
wsp = Workspace(infertiss=True, inferbat=True)
wsp.pgm = Image(np.random.rand(5, 5, 5))
wsp.pwm = Image(np.random.rand(5, 5, 5))
steps = basil.basil_steps(wsp, img)
assert (len(steps) == 3)
options = _get_defaults(img)
# options.update({
# "incpve" : True,
# })
_check_step(steps[0], desc_text="tissue", options=options)
# _check_step(steps[1], desc_text="PVE", options=options)
options.update({
"method": "spatialvb",
"param-spatial-priors": "N+",
"PSP_byname1": "ftiss",
"PSP_byname1_type": "M",
"max-iterations": 200,
"convergence": "maxits",
})
options.pop("max-trials")
_check_step(steps[2], desc_text="spatial")
def test_create_data_singleti():
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], iaf="tc", order="lrt")
assert img.ntis == 1
assert img.tis == [1.5]
assert not img.have_plds
assert img.rpts == [3]
assert img.order == "lrt"
def test_mean_across_repeats_var_rpts():
d = np.random.rand(5, 5, 5, 86)
#for z in range(8): d[..., z] = z
img = AslImage(name="asldata",
image=d,
tis=[1, 2, 3, 4, 5],
rpts=[6, 6, 6, 10, 15],
iaf="tc",
order='lrt')
img = img.mean_across_repeats()
assert img.ntis == 5
assert img.tis == [1, 2, 3, 4, 5]
assert img.rpts == [1, 1, 1, 1, 1]
assert img.ntc == 1
assert img.order == "rt"
data = img.nibImage.get_data()
assert list(data.shape) == [5, 5, 5, 5]
def test_create_data_numtis_multi():
d = np.random.rand(5, 5, 5, 8)
img = AslImage(name="asldata", image=d, ntis=2, iaf="tc", order="lrt")
assert img.ntis == 2
assert not img.tis
assert not img.have_plds
assert img.rpts == [2, 2]
assert img.order == "lrt"
def test_reorder_lrt_ltr_var_rpts():
"""
This reordering with variable repeats is not supported. In priciple
it is possible (TI1_R1, TI2_R1, TI2_R2, TI2_R3) but this seems unlikely
and is probably more likely an error
"""
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata",
image=d,
tis=[1, 2],
rpts=[1, 3],
iaf="tc",
order='lrt')
with pytest.raises(Exception):
img.reorder("ltr")
def test_create_data_diff():
d = np.random.rand(5, 5, 5, 8)
img = AslImage(name="asldata", image=d, tis=[1.5], iaf="diff", order='rt')
assert img.ntis == 1
assert img.tis == [1.5]
assert not img.have_plds
assert img.rpts == [8]
assert img.ntc == 1
assert img.order == "rt"
def test_reorder_lrt_rtl():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1], iaf="tc", order='lrt')
img = img.reorder("rtl")
assert img.ntis == 1
assert img.tis == [1]
assert not img.have_plds
assert img.rpts == [4]
assert img.ntc == 2
assert img.order == "rtl"
data = img.nibImage.get_data()
assert list(data.shape) == [5, 5, 5, 8]
for z in range(8):
if z < 4:
assert np.all(data[..., z] == z * 2)
else:
assert np.all(data[..., z] == (z - 4) * 2 + 1)
def test_infer_nothing():
"""
Check we get an error if there is nothing to infer
"""
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], order="prt")
wsp = Workspace()
with pytest.raises(ValueError):
steps = basil.basil_steps(wsp, img)
def test_calib_first_vol_fail():
""" Claim there is a calibration image as first vol but actually there isn't """
d = np.random.rand(5, 5, 5, 8)
with pytest.raises(Exception):
img = AslImage(name="asldata",
image=d,
tis=[1.5, 2.0],
iaf="tc",
order="lrt",
calib_first_vol=True)
def test_reorder_tc_ct():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1], iaf="tc", order='lrt')
img = img.reorder(iaf="ct")
assert img.ntis == 1
assert img.tis == [1]
assert not img.have_plds
assert img.rpts == [4]
assert img.ntc == 2
assert img.order == "lrt"
assert img.iaf == "ct"
data = img.nibImage.get_data()
assert list(data.shape) == [5, 5, 5, 8]
for z in range(8):
if z % 2 == 0:
assert np.all(data[..., z] == z + 1)
else:
assert np.all(data[..., z] == z - 1)
def test_pvc_only_one_map_given2():
"""
Check that PVC correction fails if you only give the WM map
"""
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], order="prt")
wsp = Workspace(infertiss=True, inferbat=True)
wsp.pwm = Image(np.random.rand(5, 5, 5))
with pytest.raises(ValueError):
basil.basil_steps(wsp, img)
def test_split_epochs_rt():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], order='rt')
imgs = img.split_epochs(4)
assert len(imgs) == 2
for idx, img in enumerate(imgs):
assert img.ntis == 1
assert img.tis == [1 + idx]
assert not img.have_plds
assert img.rpts == [1]
assert img.ntc == 1
assert img.order == "rt"
data = img.nibImage.get_data()
# Epoch 1 is TIs 1111, data 0123, mean across repeats 1.5
# Epoch 2 is TIs 2222, data 4567, mean across repeats 5.5
start = idx * 4
for z in range(data.shape[3]):
assert np.all(data[..., z] == start + z + 1.5)
def test_split_epochs_reorder():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], order='rt')
imgs = img.split_epochs(4, time_order="tr")
assert len(imgs) == 2
for idx, img in enumerate(imgs):
assert img.ntis == 2
assert img.tis == [1, 2]
assert not img.have_plds
assert img.rpts == [1, 1]
assert img.ntc == 1
assert img.order == "tr"
data = img.nibImage.get_data()
# reordered = [0, 4, 1, 5, 2, 6, 3, 7]
# Epoch 1 is TIs 1212, data 0415, mean across repeats 0.5, 4.5
# Epoch 2 is TIs 1212, data 2637, mean across repeats 2.5, 6.5
for z in range(data.shape[3]):
assert np.all(data[..., z] == idx * 2 + 0.5 + z * 4)
def test_get_nativeref_calib():
"""
Test brain extracted calibration image is used for differenced data
"""
wsp = get_wsp()
wsp.asldata = AslImage(np.random.rand(5, 5, 5, 4),
tis=[1, 2],
iaf="diff",
ibf="rpt")
reg.get_nativeref(wsp)
calib_brain = brain.brain(wsp, wsp.calib, thresh=0.2)
assert (np.allclose(calib_brain.data, wsp.reg.nativeref.data))
def test_get_nativeref_asldata_mean_ct():
"""
Test brain extracted ASL mean is used for CT data
"""
wsp = get_wsp()
wsp.asldata = AslImage(np.random.rand(5, 5, 5, 4),
tis=[1, 2],
iaf="ct",
ibf="rpt")
reg.get_nativeref(wsp)
meanasl_brain = brain.brain(wsp, wsp.asldata.mean(), thresh=0.2)
assert (np.allclose(meanasl_brain.data, wsp.reg.nativeref.data))
def test_preproc_none():
wsp = Workspace()
d = np.random.rand(5, 5, 5, 6)
wsp.asldata = AslImage(d, name="asldata", tis=[1.5], iaf="tc", order="lrt")
preproc.preprocess(wsp)
assert wsp.asldata.ntis == wsp.asldata_preproc.ntis
assert wsp.asldata.tis == wsp.asldata_preproc.tis
assert wsp.asldata.have_plds == wsp.asldata_preproc.have_plds
assert wsp.asldata.rpts == wsp.asldata_preproc.rpts
assert wsp.asldata.ntc == wsp.asldata_preproc.ntc
assert wsp.asldata.order == wsp.asldata_preproc.order
def test_get_regfrom_asldata_mean_tc():
"""
Test brain extracted ASL mean is used for TC data
"""
wsp = get_wsp()
wsp.asldata = AslImage(np.random.rand(5, 5, 5, 4),
tis=[1, 2],
iaf="tc",
ibf="rpt")
reg.get_regfrom(wsp)
meanasl_brain = brain.brain(wsp, wsp.asldata.mean(), thresh=0.2)
assert (np.allclose(meanasl_brain.data, wsp.reg.regfrom.data))
def test_split_epochs_overlap():
d = np.zeros([5, 5, 5, 8])
for z in range(8):
d[..., z] = z
img = AslImage(name="asldata", image=d, tis=[1, 2], order='tr')
imgs = img.split_epochs(4, overlap=2)
assert len(imgs) == 3
for idx, img in enumerate(imgs):
assert img.ntis == 2
assert img.tis == [1, 2]
assert not img.have_plds
assert img.rpts == [1, 1]
assert img.ntc == 1
assert img.order == "tr"
data = img.nibImage.get_data()
# Epoch 1 is TIs 1212, data 0123, mean across repeats 12
# Epoch 2 is TIs 1212, data 2345, mean across repeats 34
# Epoch 3 is TIs 1212, data 4567, mean across repeats 56
start = idx * 2
for z in range(data.shape[3]):
assert np.all(data[..., z] == start + z + 1)
def test_pvc_no_tissue():
"""
Check that PVC correction fails if you do not infer the tissue component
"""
d = np.random.rand(5, 5, 5, 6)
img = AslImage(name="asldata", image=d, tis=[1.5], order="prt")
wsp = Workspace(infertiss=False, inferbat=True)
wsp.pgm = Image(np.random.rand(5, 5, 5))
wsp.pwm = Image(np.random.rand(5, 5, 5))
with pytest.raises(ValueError):
basil.basil_steps(wsp, img)
def test_create_data_multiti_var_repeats():
d = np.random.rand(5, 5, 5, 8)
img = AslImage(name="asldata",
image=d,
tis=[1.5, 2.0],
iaf="tc",
order="lrt",
rpts=[1, 3])
assert img.ntis == 2
assert img.tis == [1.5, 2.0]
assert not img.have_plds
assert img.rpts == [1, 3]
assert img.order == "lrt"