Skip to main content
. 2021 Jun 10;12:666875. doi: 10.3389/fneur.2021.666875

Table 1.

Summary of the main article cited in this review and their main properties.

Reference Taska Input Datab Output Datac Nb subjects Data selectiond Algorithm typee Validationf Evaluation metricg Performance Model available
1) MRC CRASH Trial Collaborators (18) Cla Clinical data + RR dGOS 18517 GCS ≤ 14 MLR External AUC 77% Yes
2) Steyerberg et al. (19) Cla Clinical data + RR dGOS 14781 GCS ≤ 12 MLR External AUC 80% Yes
3) Raj et al. (9) Cla RR dGOS 869 Severe + moderate + mild complicated TBI MLR Internal AUC 75% No
4) Matsuo et al. (20) Cla Clinical data + RR dGOS 232 Abnormal RR RF Internal AUC 89.5% No
5) Hale et al. (21) Cla Clinical data + RR dGOS 565 Mild + severe pediatric TBI ANN Internal AUC 94.6% No
6) Rau et al. (22) Cla Clinical data + RR Mortality 2059 AIS≥3 MLR Internal Acc 93.5% No
7) van der Ploeg et al. (23) Cla Clinical data + RR Mortality 11026 Moderate + severe TBI MLR External AUC 76.4% No
8) Gravesteijn et al. (24) Cla Clinical data + RR Mortality 12576 Moderate + severe TBI GBM External AUC 83% No
8) Gravesteijn et al. (24) Cla Clinical data + RR dGOS 12576 Moderate + severe TBI ANN External AUC 78% No
9) Kim et al. (25) Cla CT-scan Severe/mild edema 70 Pediatric TBI Proportion of voxels ∈[17, 24] HU + non parametric tests NI AUC 85% No
9) Kim et al. (25) Cla CT-scan Delayed/mild edema 70 Pediatric TBI Proportion of voxels ∈[17, 24] HU + non parametric tests NI AUC 75% No
10) Rosa et al. (17) Cla CT-scan + lesions segmentation EDH + SDH + Contusions 155 Presence lesion Radiomic features extraction + PLS-DA Internal Acc 89.7% No
11) Chilamkurthy et al. (26) Cla CT-scan ICH + fracture + midline shift + mass effect 313809 NI CNN External AUC 92.16 - 97.31% No
12) Jadon et al. (27) Seg 2D CT-scan Hemmorhage 40000 NI CNN NI DSC 85.78 - 94.24% No
13) Jain et al. (28) Seg CT-scan IC lesions 144 Center-TBI CNN Internal DSC 73% No
14) Kuo et al. (29) Seg CT-scan ICH 791 NI CNN External DSC 76.6% No
15) Yao et al. (30) Seg CT-scan Hematoma 828 GCS∈[4, 12] CNN Internal DSC 69.7% No
15) Yao et al. (30) Cla Clinical data + CT-scan Mortality 828 GCS∈[4, 12] RF Internal AUC 85.3% No
16) Monteiro et al. (16) Seg CT-scan IPH + EAH + PO + IVH 839 Center-TBI CNN Internal DSC 36% Yes
16) Monteiro et al. (16) Cla CT-scan IPH + EAH + PO + IVH 490 Center-TBI + CQ500 CNN External AUC 83% - 95% Yes
a

Task: Cla, Classification; Seg, Segmentation.

b

Input Data: clinical data = metrics representing demography or physiology, RR, radiological reading metrics manually retrieved from CT scan and CT scan, computed tomography image.

c

Output Data: dGOS, dichotomized Glasgow Outcome Score, EDH, extra dural hemmorhage; SDH, subdural hemorrhage; ICH, intracranial hemorrhage; IC, intracranial; PO, oerilesional edema; IVH, intraventricular hemorrhage.

d

Data selection: GCS, Glasgow Coma Score; AIS, Abbreviated Injury Scale; NI, no information; Center-TBI and CQ500: public databases containing TBI CT scans.

e

Algorithm type: MLR, multivariate logical regression; RF, random forest; ANN, artificial neural network; CNN, convolutional neural network; GBM, gradient boosting machine; HU, Hounsfield Units.

f

Validation: NI, no information.

g

Evaluation metric: AUC, area under the curve; Acc, accuracy; DSC, Dice similarity coefficient.