fidopt Interactive view
MR to MR, another contrast, field strength or scanner, synthesis

Duke breast DCE MR, pre-contrast to first post-contrast

Fat-saturated T1-weighted breast MR of a woman with invasive breast cancer before surgery: the pre-contrast series (source) and the first post-contrast series (target) of one dynamic examination, on one grid (Duke-Breast-Cancer-MRI, TCIA).

30 test casesbreastsource T1 pre-contrast, target T1 first post-contrastresearch only
Look at the images
Read before comparing (8)
  • Pool. Duke-Breast-Cancer-MRI holds the pre-operative DCE MR of 922 women with biopsy-confirmed invasive breast cancer, imaged prone at 1.5 or 3 T on GE and Siemens scanners at one hospital over a decade. A case is a patient's examination; the test cases come from six scanner models, SIGNA HDx the most frequent.
  • Pair. Source and target are the series the provider's mapping table names pre and post_1: the fat-saturated T1-weighted gradient echo before contrast and its first repetition after. They are separate series of one examination with one geometry, checked case by case: the target must lie on the source's grid, or the case stops. Nothing is registered, and motion between the two series counts as error for every method: a rigid fit, not applied, moves the target by 0.29 mm in the median over the 30 test cases, over 1 mm on 4 of them and 3.7 mm on Breast_MRI_387.
  • Grid. Breast_MRI_679 re-centred its field by 16.8 mm between the two series, so its target does not lie on the source's grid: it is held out, and Breast_MRI_139 joins development in its place.
  • Images. Axial, 0.57 to 1.19 mm in plane and 1 to 2.5 mm between slices, as delivered.
  • Values. The scanners' magnitudes without unit, on scales that differ by scanner: the target's 99.9th percentile inside the mask runs from 301 to 5,232 on the test cases. truth.dynamic_range, 3,978, is that percentile over every voxel inside the masks of the 30 test cases: the PSNR and SSIM of dim scans read high, and the MAE weighs the bright ones most.
  • Mask. The body in the field, from the source alone by an Otsu threshold on its log intensities: the breasts, the chest wall and part of the thorax, 32% to 70% of the image on the test cases; on some cases (Breast_MRI_066) it takes in a band of the air's noise along the skin.
  • Baseline. Inside the mask on the 30 test cases a constant at the target's median scores an MAE of 237 (28 to 602), a PSNR of 24.1 dB (12.7 to 39.2) and an SSIM of 0.49 (0.16 to 0.88). The source itself scores 162 (14 to 403), 26.7 dB and 0.64, better than the constant on all 30: what a method must add is the enhancement.
  • Overlap. The 209 patients with one of the provider's breast and fibroglandular tissue segmentations (the SEG series and the 3D supplement) are kept out of the pool for DukeBreastFGT, which takes those images. MAMA-MIA, outside the corpus, holds Duke patients: a network trained on it may have seen the test cases.

No method has a result on every case of this dataset yet.

Case by case

Click a score to look at that result: the target image, and over it the method's synthetic image. Underlined scores have their image; for the others you see the case without the result.

About the data

Test cases
30 test patients drawn with seed 2026 by scanner maker in proportion (20 GE, 10 Siemens) among the 713 of the 922 without one of the provider's breast tissue segmentations, which a segmentation dataset of these images takes, one examination each. Then 30 development patients, the last of the same seeded order, each maker in proportion: 20 GE, 10 Siemens. Breast_MRI_679, drawn for development, is held out (draw.held_out): its post-contrast series lies 16.8 mm off its pre-contrast grid.
Truth
target image as released the patient's first post-contrast series, as the provider sorted the examination's series (post_1), a series of the same examination and geometry as the source, on its grid as acquired, not registered
Licence
CC BY-NC 4.0
30 cases for tuning, never tested

Breast_MRI_744, Breast_MRI_698, Breast_MRI_518, Breast_MRI_871, Breast_MRI_042, Breast_MRI_715, Breast_MRI_609, Breast_MRI_040, Breast_MRI_540, Breast_MRI_380, Breast_MRI_526, Breast_MRI_122, Breast_MRI_407, Breast_MRI_364, Breast_MRI_211, Breast_MRI_535, Breast_MRI_455, Breast_MRI_912, Breast_MRI_514, Breast_MRI_327, Breast_MRI_167, Breast_MRI_761, Breast_MRI_591, Breast_MRI_676, Breast_MRI_445, Breast_MRI_120, Breast_MRI_869, Breast_MRI_845, Breast_MRI_645, Breast_MRI_139