Table 1.
The datasets utilized for model training and validation consisted of both pediatric (denoted with suffix p) and adult (denoted with suffix a) data.
| Scenario | Dataset | Subjects | Age range (years) | Diagnosis | Female (%) | Scanner type | Source |
|---|---|---|---|---|---|---|---|
| Training | 1.1.p | 157 | 5–21 | DD and HC | 40% | 3T: Siemens | CMI-HBN17 |
| 1.2.p | 66 | 2–6 | HC | 48% | 3T: GE | Calgary18 | |
| 1.3.a | 32 | Adults | VLOSLP, LOD and HC | NA | 3T: Philips | Research cohort19 | |
| 1.4.a | 135 | 16–81 | MS | 60% | 1.5–3T: Siemens, GE and Philips, 1.5T: Fujifilm | Clinical practice | |
| Accuracy | 2.p | 103 | 4–16 | BD, Sz and HC | 45% | 1.5T: GE | CANDIShare20 |
| 2.a | 30 | 18–90 | HC | 66% | 1.5T: Siemens | MICCAI201221 | |
| Reproducibility | 3.p | 70 | 6–17 | DD and HC | 44% | 3T: Siemens | NKI22 |
| 3.a | 10 | 39–57 | MS | 70% | 3T: Siemens, GE and Philips | Re3T7 | |
| Diagnosis Performance | 4.p | 21 | 4–13 | CVI and no CVI | 23% | 1.5T:Siemens and Philips, 3T: Philips | Clinical practice |
| 4.a | 46 | 58–85 | AD and HC | 54% | 1.5T: GE | MIRIAD23 |
A subset of patients were randomly selected from original training datasets. These datasets include individuals with Developmental Disorders (DD), Healthy Control (HC), Very-Late-Onset Schizophrenia-Like Psychosis (VLOSLP), Late-Onset Depression (LOD), Bipolar Disorder (BD), Schizophrenia (Sz) and Multiple Sclerosis (MS), Cerebral Visual Impairment (CVI) and Alzheimer’s Disease (AD). NA = not available.