def test_log_path():
log.set_mode(log.Mode.LOG)
log.set_path("check/")
img = cv2.imread("tests/data/orange.png")
log.image(log.Level.ERROR, img)
assert os.path.exists('check/cvlog.html') is True
log.set_path("log/")
def localise_phantom(im, threshold = None):
"""
Show the position of a phantom in an image. Originally written for
circular phantoms. Works in 3d, expects dim order z, y, x
"""
if not threshold:
threshold = im.mean()
seg = im > threshold
seg = binary_fill_holes(seg)
seg = find_largest_segmented_object(seg)
output = localise_segmented_object(seg)
center = output['center']
#Create a debug image
logging=False
if logging:
for x in output['x_range']:
seg[:,x-2:x+2] = 1
for y in output['y_range']:
seg[y-2:y+2,:] = 1
seg[center[0]-20:center[0]+20,center[1]-5:center[1]+5] = 0
seg[center[0]-5:center[0]+5,center[1]-20:center[1]+20] = 0
log.image(log.Level.INFO, seg*.99)
return output
def test_log_image():
remove_dirs('log/')
img = cv2.imread("tests/data/orange.png")
log.set_mode(log.Mode.LOG)
log.image(log.Level.ERROR, img)
logitem = get_html('log/cvlog.html').select('.log-list .log-item')
assert logitem[0].select('.log-type')[0].text == 'image'
assert logitem[0]['logdata'] == read_file('tests/data/expected/image.txt')
def test_log_interruption():
remove_dirs('log/')
log.set_mode(log.Mode.LOG)
img = cv2.imread("tests/data/orange.png")
log.image(log.Level.ERROR, img)
assert os.path.exists('log/cvlog.html') is True
remove_dirs('log/')
log.image(log.Level.ERROR, img)
assert os.path.exists('log/cvlog.html') is True
def test_message():
remove_dirs('log/')
img = cv2.imread('tests/data/contour.jpg')
log.set_mode(log.Mode.LOG)
imgray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret, thresh = cv2.threshold(imgray, 127, 255, 0)
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
message = 'Lorem ipsum dolor sit amet, ne persius reprehendunt mei. Ea summo elitr munere his, et consul offendit recteque sea, quis elit nam ut.'
log.image(log.Level.ERROR, img)
log.contours(log.Level.ERROR, contours, img, msg=message)
logitem = get_html('log/cvlog.html').select('.log-list .log-item')
assert logitem[0].select('.description') == []
assert logitem[1].select('.description')[0].text == message
def calculate_MTF_from_ROI(im_ROI):
mask = iu.mask_point(im_ROI)
mask = morphology.binary_opening(mask)
log.image(log.Level.TRACE, im_ROI * mask)
mask = morphology.binary_dilation(mask, iterations = 3)
im_ROI = subtract_background(im_ROI, mask)
localisation = iu.localise_segmented_object(mask)
y, x = localisation['center']
y_vals = iu.get_nearby_integers(y)
x_vals = iu.get_nearby_integers(x)
log_text(x_vals, priority = 5)
log_text(y_vals, priority = 5)
h_LSF = make_LSF(im_ROI, y_vals, 1)
v_LSF = make_LSF(im_ROI, x_vals, 0)
h_MTF = iu.mtf_from_lsf(h_LSF)
v_MTF = iu.mtf_from_lsf(v_LSF)
return h_MTF, v_MTF
def catphan_select_wire_roi(im):
"""
Take an image of the contrast segment of a catphan. If the phantom
is correctly oriented, the wire will be in a similar position relative to
the phantom boundaries.
"""
#Based on phantom measurements made in ImageJ
#y_min, y_max = 18.1, 219
#x_min, x_max = 16.9, 217.2
#w_x, w_y = 140.5, 120.4
#yratio = (w_y-y_min)/(y_max - y_min)
#xratio = (w_x-x_min)/(x_max - x_min)
yratio = 0.5092085614733699
xratio = 0.6170743884173739
#Threshold the phantom out
p_shape = im > -900
log.image(log.Level.TRACE, im * p_shape)
p_shape = morphology.binary_erosion(p_shape, iterations = 12)
p_shape = morphology.binary_dilation(p_shape, iterations = 12)
log.image(log.Level.TRACE, im * p_shape)
p_y, p_x = np.where(p_shape==True)
p_y.min(), p_x.max()
#Select the position of the wire based on previous measurements
ypos = yratio * (p_y.max() - p_y.min()) + p_y.min()
xpos = xratio * (p_x.max() - p_x.min()) + p_x.min()
r = np.array([int(xpos)-14, int(ypos)-14,28,28])
a = im[iu.Rslice(r)].copy()
log.image(log.Level.INFO, a)
return a
def align_line_profiles(ROI, midpoint=None):
"""
Align a set of line profiles arranged in columns, centering at the half-max
Parameters
----------
ROI : np.array
ROI containing line profiles.
Returns
-------
output : np.array
ROI with profiles aligned at the half-maximum.
"""
if not midpoint:
midpoint = (ROI.max() + ROI.min())/2
smoothed = np.apply_along_axis(wiener, 0, ROI)
indexes = np.apply_along_axis(interp_first_index, 0, smoothed, midpoint)
shifts = ROI.shape[0]//2 - indexes
output = shift_columns_by_subpixel(ROI, shifts)
log.image(log.Level.TRACE, apply_window(output))
return output
def log_all_level(img):
log.image(log.Level.TRACE, img)
log.image(log.Level.INFO, img)
log.image(log.Level.ERROR, img)
def old_clinical_mtf(im, pixel_spacing, roi_size=(50, 30)):
"""
Calculate the MTF from a patient/phantom image.
Requires air on the side of
the patient corresponding to the zero direction of the y axis.
Only use this function for 2d images, which are not currently covered by
new clinical MTF
Parameters
----------
im : np.array
2d or 3d axial CT image. im[Z, Y, X]
pixel_spacing : tuple
pixel size of (y, x) in mm.
roi_size : tuple,
Size of the ROI that the MTF will be calculated across. The default is (50, 30).
Returns
-------
output : dictionary
Dictionary containing the key 'clinical', which contains 'MTF' and 'frequency'
"""
if im.ndim == 2:
im = im[np.newaxis, ]
mtfs = []
fs = []
for i in range(im.shape[0]):
im_2d = im[i, :]
#find head
try:
seg_data = iu.localise_phantom(im_2d, -300)
if seg_data['anterior_point'][0] < 15:
im_2d = np.pad(im_2d, 15, mode='constant',
constant_values=-1024)
seg_data = seg_data = iu.localise_phantom(im_2d, -300)
except:
continue
y = seg_data['anterior_point'][0]
x = seg_data['anterior_point'][1]
roi = iu.extract_patch_around_point(im_2d, (y, x), roi_size)
log.image(log.Level.TRACE, iu.apply_window(roi))
# find the index of array element with the highest rate of change
roi = iu.align_line_profiles(roi, -500)
w = 10
c = roi.shape[0]//2
esf = roi[c-w:c+w, :].mean(axis=1)
#log_plot(range(len(esf)), esf)
mtf = iu.mtf_from_esf(esf)
mtf = mtf/mtf[0]
f = pixel_spacing[0]/2*np.arange(mtf.shape[0])
mtfs.append(mtf)
fs.append(f)
mtf = np.array(mtfs)
mtf = np.median(mtf, axis=0)
f = np.array(fs)
f = np.median(f, axis=0)
output = {'Clinical': {'MTF': mtf,
'frequency': f}}
return output
