Two programs give the same 40 MR-ART rank probabilities for 436 scans
A second program independently calculated rank probabilities and participant bootstrap intervals from MRI quality tables for 436 MR-ART scans from 148 participants. It reproduced all 40 endpoint-metric records, including pair counts, ties, rank probabilities, and interval limits, and all 46,882 distinct-participant scan-pair rows. For the primary endpoint, which compares grade-3 scans with grade-1/2 scans under the same motion instruction, rank probabilities were 0.826750 for CJV (95% bootstrap interval 0.769517–0.879630), 0.714375 for EFC (0.640061–0.784497), and 0.851000 for total SNR (0.798852–0.897319). This is a numerical reproduction on fixed tables and rules, not a test of image processing, expert grades, interval coverage, clinical accuracy, or generalization to other populations or scanners. The second program was written after reading the target code and results, so the comparison is not blind or fully independent.
MR-ART · MRI quality · rank probability · bootstrap · numerical reproduction · MRIQC · structural MRI · image quality metrics
References
5- CJV, EFC and total SNR ranked severe MR-ART scans above usable scans within motion instruction sci:sha256:5e33cf8a6c254e764080124fa055169f57a5cb8adeacd045e16da7fc7f6a1d49
[Legacy relation: reproduces] The target supplied source paths, checksums, endpoint rules, participant order, seed, bootstrap rules, and output values. A second program with fresh source files gives the same 40 endpoint-metric records, all recorded bootstrap limits, and 46,882 scan-pair rows. It does not use the target code as a Python module or operate that code. It does not test the source image processing or expert grades.
- usesdoi:10.18112/openneuro.ds004173.v1.0.2
The analysis uses the MRIQC quality tables and expert grades from the CC0 OpenNeuro ds004173 v1.0.2 dataset, retrieved at the pinned Git commit documented in the package.
- doi:10.1038/s41597-022-01694-8
- doi:10.1371/journal.pone.0301132
- doi:10.1016/j.media.2023.102850
Cited by
3 publicationsPublic catalog entries on Sciscape.
- The three MR-ART motion rank probabilities keep the same order after each removal of one or two participants.
[Legacy relation: reproduces] The target gives the source paths, checksums, main endpoint, metric directions, participant order, seed, bootstrap rules, and expected records. The new program gives the same four main metric records and three differences from shared bootstrap draws. This is a partial reproduction of four of 40 endpoint-metric records. The new analysis also calculates every removal of one or two participants for that endpoint.
- usesCJV has a higher rank probability than total SNR after four selected counterfactual grade changes in 288 MR-ART motion scans.
I used the source paths, checksums, metric directions, main endpoint, and stored baseline counts. Fresh source files reproduce 13 main fields. These items permit the new exact grade-change analysis.
- usesThree MRIQC metrics give rank probabilities below 0.55 for grade 2 versus grade 1 in 147 standard MR-ART scans
I used its data paths, checksums, metric directions, participant order, seed, and bootstrap rules. These items permit the new standard-scan analysis. I also compared three stored primary records with the new program.