Harmonization of diffusion MRI measures is crucial for white matter tract normative assessment in ADNI
Bibliographic record
Abstract
Abstract Background Diffusion MRI (dMRI) measures are variable across sites and MRI vendors, which leads to a site bias in the reported quantitative white matter metrics (Figure 1). ComBat is currently the go‐to method for harmonizing MRI data. However, to our knowledge, the harmonization power of ComBat has not been convincingly demonstrated on the ADNI cohort and was never tested in the context of normative assessment for Alzheimer’s disease (AD) patients. Methods We select 21 sites in the ADNI cohort with at least 10 healthy controls (HC) datapoints per site. Instead of harmonizing all patients from every site onto an average template, as done in ComBat, we identified a reference site containing a large number of HC of all ages and gender. We use the Cam‐CAN dataset, with 441 subjects aged 18‐87. Moreover, instead of computing the harmonization function from every subject (HC and AD), only HC are used. The harmonization function of each site is thus computed in a pairwise fashion between each ADNI site and the Cam‐CAN site. Finally, once the harmonization function of each site is computed, all subjects of every site are harmonized towards the Cam‐CAN site. All results are shown for the Mean Diffusivity and Apparent Fiber Density dMRI measures in the left arcuate fasciculus. Results The left panel of Figure 1 shows how dMRI measures vary across sites/vendors. The most striking bias is between Philips sites/scanners and the rest. The right panel of Figure 1 shows the successful harmonization effect of ComBat. In Figure 2, the Cam‐CAN reference is shown in gray with different percentile distributions across ages and AD patients from the GE ADNI site #127 before and after harmonization. Harmonization clearly moves HC and AD datapoints in the correct measurement “space” and preserves disease stage (HC vs AD). Conclusions We showed that harmonization is crucial in white matter dMRI in ADNI. Harmonization methods can correct site biases and are robust to pathology, as HC and AD patients are successfully harmonized across sites. Harmonization is thus crucial for data pooling and quantitative comparison in large multi‐site datasets such as ADNI.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.023 | 0.068 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.002 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".