| Autor: |
Jenna M. Schabdach, J. Eric Schmitt, Susan Sotardi, Arastoo Vossough, Savvas Andronikou, Timothy P. Roberts, Hao Huang, Viveknarayanan Padmanabhan, Alfredo Oritz-Rosa, Margaret Gardner, Sydney Covitz, Saashi A. Bedford, Ayan Mandal, Barbara H. Chaiyachati, Simon R. White, Ed Bullmore, Richard A.I. Bethlehem, Russell T. Shinohara, Benjamin Billot, J. Eugenio Iglesias, Satrajit Ghosh, Raquel E. Gur, Theodore D. Satterthwaite, David Roalf, Jakob Seidlitz, Aaron Alexander-Bloch |
| Rok vydání: |
2023 |
| DOI: |
10.1101/2023.01.13.23284533 |
| Popis: |
BackgroundBrain MRIs acquired in clinical settings represent a valuable and underutilized scientific resource for investigating neurodevelopment. Utilization of these clinical scans has been limited because of their clinical acquisition and technical heterogeneity. These barriers have curtailed the interpretability and scientific value of retrospective studies of clinically acquired brain MRIs, compared to studies of prospectively acquired research quality brain MRIs.PurposeTo develop a scalable and rigorous approach to generate clinical brain growth chart models, to benchmark neuroanatomical differences in clinical MRIs, and to validate clinically-derived brain growth charts against those derived from large-scale research studies.Materials and MethodsWe curated a set of clinical MRIScans withLimitedImagingPathology (SLIP) – so-called “clinical controls” – from an urban pediatric healthcare system acquired between 2005 and 2020. The curation process included manual review of signed radiology reports, as well as automated and manual quality review of images without gross pathology. We measured global and regional volumetric imaging phenotypes in the SLIP sample using two alternative, advanced image processing pipelines, and quantitatively compared clinical brain growth charts to research brain growth charts derived from >123,000 MRIs.ResultsThe curated SLIP dataset included 372 patients scanned between the ages of 28 days post-birth and 22.2 years across nine 3T MRI scanners. Clinical brain growth charts were highly similar to growth charts derived from large-scale research datasets, in terms of the normative developmental trajectories predicted by the models. The clinical indication of the scans did not significantly bias the output of clinical brain charts. Tens of thousands of additional healthcare system scans meet inclusion criteria to be included in future brain growth charts.ConclusionBrain charts derived from clinical-controls are highly similar to brain charts from research-controls, suggesting that curated clinical scans could be used to supplement research datasets.Summary StatementBrain growth charts of pediatric clinical MRIs with limited imaging pathology (N=372) are highly correlated with charts from a large aggregated set of research controls (N>120,000).Key ResultsA cohort of brain MRI scans with limited reported imaging pathology (N=372, 186 female; ages 0.07 - 22.2 years, median = 10.2) were identified using signed radiology reports and processed using two segmentation pipelines. Growth charts generated from these scans are highly correlated with growth charts from a large aggregated set of research controls (r range 0.990 - 0.999). There was no evidence of bias due to the reason for each scan. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
