STATISTICS.md

======================================================================== HEALTHY-BRAIN TRAINING CORPUS — FINAL SNAPSHOT

Manifest: /home/MRI-DataSet/manifest.csv Golden-0-to-25 (primary): /home/MRI-DataSet/Golden-0-to-25/ Golden-25plus (fine-tune): /home/MRI-DataSet/Golden-25plus/

Total scans: 6351 Labelled healthy: 4592 (72.3%) With known age: 6268 (98.7%) age < 25 (primary): 4782 age >= 25 (pretrain): 1486 Unknown age (to impute): 83

Age distribution (years) over dated scans: min=-0.1 median=15.2 max=86.3 mean=20.3 ± 17.6

---

SPLIT BREAKDOWN

0-25 4782 scans (healthy=3141) range=-0.1-25.0 25+ 1370 scans (healthy=1370) range=25.0-86.3 excluded_patho_over_25 118 scans (healthy=0) range=25.0-64.0 unknown 81 scans (healthy=81) range=n/a

---

PER-DATASET BREAKDOWN

dataset scans healthy age<25 age>=25 age range

DataSet-10_ABIDE-I 1102 571 941 161 6.5-64.0 DataSet-11_ABIDE-II 1007 533 901 106 5.1-64.0 DataSet-12_ADHD-200 717 230 717 0 7.2-21.7 DataSet-1_BCP 566 566 566 0 -0.1-3.0 DataSet-1_BCP_patho 267 0 267 0 -0.1-3.0 DataSet-2_Calgary 266 266 223 0 2.6-6.9 DataSet-3_ds002726 29 29 29 0 10.0-10.0 DataSet-4_ds000248 1 1 0 1 30.0-30.0 DataSet-5_PTBP 213 213 193 0 7.0-18.0 DataSet-6_IXI 581 581 42 521 20.0-86.3 DataSet-7_MPI-Leipzig 319 319 105 212 22.5-77.5 DataSet-8_AOMIC-ID1000 951 951 798 153 19.0-26.0 DataSet-9_NKI-Rockland 332 332 0 332 38.0-74.0

AGE HISTOGRAM (5-year bins, across all dated scans) [ 0, 5) 985 ############################### [ 5, 10) 853 ########################### [ 10, 15) 1242 ######################################## [ 15, 20) 536 ################# [ 20, 25) 1161 ##################################### [ 25, 30) 418 ############# [ 30, 35) 136 #### [ 35, 40) 86 ## [ 40, 45) 108 ### [ 45, 50) 111 ### [ 50, 55) 98 ### [ 55, 60) 133 #### [ 60, 65) 132 #### [ 65, 70) 129 #### [ 70, 75) 112 ### [ 75, 80) 15
[ 80, 85) 6
[ 85, 90) 2

======================================================================== TARGET SPLITS FOR TRAINING

Primary training (0-25 years) : 4782 scans Pretraining pool (25+ years) : 1486 scans Grand total labelled : 6268 scans

Strategy (per user):

1. Train EfficientNet-B3 multi-view on Training_0to25/ (pediatric + young adult)
2. Ablation: pretrain on Training_0to25 + Pretrain_25plus (full lifespan), then fine-tune on 0-25
3. Compare pediatric MAE between the two strategies in thesis