ACR v1.1 - Refactor and Improved Masking

[YassineAzma]

This PR introduces the ACRTools class in acr_tools.py, which contains a number of useful functions that are frequently used throughout the ACR analysis scripts.

sort_images = DICOM and image sorting via the image position

determine_rotation = Estimate the rotation angle of the phantom using the Hough circle detector

rotate_images = Rotates the images by the rotation angle

mask_image = Thresholds the image - if the intensity variation in the image is sufficiently large, local thresholding methods are used rather than a global intensity threshold.

find_phantom_center = Now decoupled from the masking routine. The image is blurred and the edges are then sharpened along the diagonal. The Hough circle detector is then used to identify the phantom using prior knowledge of its geometry.

circular_mask = Draw a circular mask

measure_orthogonal_lengths = Measure the horizontal and vertical extent of the mask.

rotate_point = Find x' and y' for a point rotated around a given centre by a given angle.

find_n_highest_peaks = Given an array, report the location and intensity of the highest peaks.

Along with other general improvements, these functions have been used to refactor the existing ACR functions. This will prevent unneeded repetition in the unit tests for the individual analysis scripts by grouping tests for reused functions in the acr_tools.py module.

NOTE: All ACR test outputs have changed by <2% owing to changes in the masking and centre definition methods. These changes should make them more robust to issues seen in Issues #333 and #332.

Pending: Complete unit tests for acr_tools module.

Resolves #333, #332

[github-actions[bot]]

Coverage Report

File	Stmts	Miss	Cover	Missing
hazenlib
ACRObject.py	83	3	96%	66, 82–83
HazenTask.py	26	3	88%	32–34
__init__.py	54	17	69%	125–133, 145, 179–181, 184–186, 193–196, 207
exceptions.py	21	4	81%	17–21
relaxometry.py	286	86	70%	182–200, 377, 423–425, 492, 566–588, 606–621, 1037–1040, 1046–1052, 1085–1130
utils.py	189	43	77%	61, 65, 75, 80, 117, 124–129, 140, 143–150, 170–172, 190–192, 211–213, 222, 227, 233, 284, 287, 295–300, 303, 346, 355, 371
hazenlib/tasks
acr_geometric_accuracy.py	121	64	47%	41–73, 121–146, 160–194
acr_ghosting.py	109	43	61%	34–49, 75–77, 107–109, 145–186
acr_slice_position.py	135	47	65%	46–61, 187–233
acr_slice_thickness.py	134	58	57%	34–48, 157–216
acr_snr.py	135	63	53%	43–83, 93, 163–173, 207–222, 255–272
acr_spatial_resolution.py	209	71	66%	58–83, 128, 171, 184–193, 275–326
acr_uniformity.py	83	33	60%	35–51, 108–133
ghosting.py	152	53	65%	18–35, 50, 112–113, 117, 127–128, 154–156, 173–175, 221–259
relaxometry.py	7	7	0%	1–11
slice_position.py	117	23	80%	28, 37–38, 49, 103–104, 130, 210, 217–234
slice_width.py	360	53	85%	34–37, 41, 109, 168–188, 453, 458–459, 465, 470, 532–533, 782–823
snr.py	166	67	60%	51, 68–73, 167–185, 200–209, 227–237, 264–274, 279–289, 320–333, 338–346, 375–388
snr_map.py	104	1	99%	291
spatial_resolution.py	248	45	82%	36–39, 43, 64, 149, 208, 334–370
uniformity.py	79	20	75%	42–45, 51, 93–94, 101, 135–149
TOTAL	2843	804	72%

Tests	Skipped	Failures	Errors	Time
201	0 :zzz:	0 :x:	0 :fire:	3m 20s :stopwatch:

[tomaroberts]

Just to add to this, you've written a fair amount of documentation in the files, which is great. When we come to fully uplifting the hazen ReadTheDocs, we'll definitely need your input on making the guidance as clear as possible. Thanks.