Graph-selected brain connectivity is associated with ADHD symptom scores
Source check: Source checked. Not independent scientific confirmation.
Finding: A dynamic graph convolutional network analyzed resting-state scans from 603 ADHD-200 participants. Model-selected connectivity features were associated with symptom severity in a 172-person subgroup. These retrospective, post-selection associations are exploratory and do not establish a causal ADHD mechanism or clinical utility.
Technology: Dynamic graph convolutional network (dGCN) and isolation forest
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Learned graph convolutions ranked regional connectivity features used to investigate symptom associations; a separate conventional linear regression quantified symptom-score prediction. The claimed result is the exploratory AI-selected association, not a causal mechanism.
- Test and result
- The selected four-site sample contained 260 ADHD and 343 control participants after isolation-forest outlier removal. dGCN combined connectivity with seven personal-characteristic variables and reached 72.0% diagnostic accuracy in site-stratified ten-fold cross-validation; leave-site-out accuracy ranged from 67.9% to 74.3%. Across 172 Peking University participants with symptom scales, regression on model-selected regional features yielded cross-validated symptom-score correlations of .33–.36. Selected within-frontal, within-posterior-fossa and subcortical–posterior-fossa connectivity showed symptom correlations of .28–.34, reported with FDR correction.
- Limitation / still unknown
- Regional selection used model weights averaged across diagnostic cross-validation. Its independence from the subsequent symptom regression and correlation analyses is not established; FDR correction does not remove selection bias. All symptom analyses used a subset of the same source dataset. Imaging alone did not supply the reported diagnostic predictions: age, sex, handedness and IQ-related inputs were included. Site-held-out classification is not independent replication of the selected biological associations. No causal direction, independent symptom-cohort confirmation or benefit from clinical use is established. The November2021 online version is represented; the issue date is February2022, and the bounded search does not establish earliest public priority.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10569447/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 21 Nov 2026. Review scope: Original publication metadata, computational Methods2.1/2.2.6, Results3.1/3.3.1–3.3.3 and limitations4.2 actually read; no author code executed..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 1/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.