Copyright
©The Author(s) 2020.
World J Gastroenterol. Jul 7, 2020; 26(25): 3660-3672
Published online Jul 7, 2020. doi: 10.3748/wjg.v26.i25.3660
Published online Jul 7, 2020. doi: 10.3748/wjg.v26.i25.3660
Table 1 The basic information and detail distribution of each dataset
Training set | Test set | P value | ||
Category A: HCC | No. of lesions/No. of patients | 88/79 | 23/22 | |
Age (median [range]) in yr | 49 (24-81) | 49.5 (33-70) | 0.726 | |
Sex (percentage of women) | 6/79 (7.6%) | 5/22 (22.7%) | 0.044 | |
Size of lesion (mean ± SD) in mm | 60.6 ± 36.3 | 63.0 ± 45.4 | 0.789 | |
Histopathologic diagnosis (No. of lesions/No. of patients) | ||||
Surgery | 79/70 | 20/19 | ||
Biopsy | 9/9 | 3/3 | ||
Category B: Metastases | No. of lesions/No. of patients | 89/34 | 23/14 | |
Age (Median [range]) (yr) | 58.5 (23-79) | 58 (23-79) | 0.937 | |
Sex (Percentage of women) | 8/34 (23.5%) | 6/14 (42.9%) | 0.181 | |
Size of lesion (mean ± SD) in mm | 23.0 ± 13.9 | 22.7 ± 11.5 | 0.937 | |
Primary tumors (No. of lesions/No. of patients) | ||||
Colorectal cancer | 40/20 | 10/6 | ||
Gastric carcinoma | 13/3 | 3/2 | ||
Breast cancer | 2/1 | 0/0 | ||
Lung cancer | 14/4 | 4/2 | ||
Thyroid cancer | 16/4 | 4/2 | ||
Malignant jejunal stromal tumor | 2/1 | 1/1 | ||
Duodenal papillary carcinoma | 0/0 | 1/1 | ||
Laryngocarcinoma | 2/1 | 0/0 | ||
Category C: Benign non-inflammatory FLLs | No. of lesions/No. of patients | 128/97 | 34/32 | |
Age (median [range]) in yr | 34 (17-82) | 34 (10-74) | 0.729 | |
Sex (percentage of women) | 52/97 (53.6%) | 16/32 (50.0%) | 0.723 | |
Size of lesion (mean ± SD) in mm | 41.9 ± 30.5 | 52.9 ± 28.4 | 0.060 | |
Histological type (No. of lesions/No. of patients) | ||||
Hemangioma | 55/35 | 15/15 | ||
FNH | 67/58 | 17/15 | ||
Adenoma | 6/4 | 2/2 | ||
Category D: Hepatic abscesses | No. of lesions/No. of patients | 105/77 | 27/20 | |
Age (median [range]) in yr | 54 (4-82) | 55 (25-82) | 0.936 | |
Sex (percentage of women) | 24/77 (31.2%) | 7/20 (35.0%) | 0.743 | |
Size of lesion (mean ± SD) in mm | 64.5 ± 34.9 | 63.8 ± 24.2 | 0.916 |
Table 2 The confusion matrix analysis on test set
Ground truth | Positive predictive value | |||||
Benign | Metastases | HCCs | Hepatic abscesses | |||
non-inflammatory FLLs | ||||||
Prediction | Benign non-inflammatory FLLs | 25 | 0 | 4 | 2 | 0.806 |
Metastases | 3 | 23 | 2 | 3 | 0.742 | |
HCCs | 3 | 0 | 17 | 0 | 0.85 | |
Hepatic abscesses | 3 | 0 | 0 | 22 | 0.88 | |
Sensitivity | 0.735 | 1 | 0.739 | 0.815 | ||
Specificity | 0.918 | 0.905 | 0.964 | 0.963 | ||
Accuracy | 0.86 | 0.925 | 0.916 | 0.925 | ||
Mean accuracy | 0.813 |
Table 3 The model's performance comparison between the normal set and “phase cheating” sets
Policy | HCCs (AUC [95%CI]/P value) | Metastases (AUC [95%CI]/P value) | Benign non-inflammatory (AUC [95%CI]/P value) | Hepatic Abscesses (AUC [95%CI]/P value) |
PP + AP + PVP + DP | 0.92 (0.837-0.992) | 0.99 (0.967-1.00) | 0.88 (0.795-0.955) | 0.96 (0.914-0.996) |
AP + AP + PVP + DP | 0.820 (0.705-0.905)/0.0699 | 0.901 (0.805- 0.960)/0.0289 | 0.893 (0.809-0.949)/0.2502 | 0.924 (0.823-0.977)/0.3387 |
PP + PVP + PVP + DP | 0.704 (0.565-0.821)/0.0017 | 0.930 (0.832- 0.981)/0.2573 | 0.799 (0.701- 0.877)/0.0924 | 0.938 (0.846-0.984)/0.4317 |
PP + AP + AP + DP | 0.768 (0.643-0.866)/0.0013 | 0.833 (0.714 -0.916)/0.0120 | 0.864 (0.774-0.929)/0.9720 | 0.935 (0.846-0.981)/0.4047 |
PP + AP+ PVP + PVP | 0.911 (0.815-0.967)/0.6404 | 0.959 (0.882- 0.992)/0.4066 | 0.913 (0.832-0.963)/0.7877 | 0.831 (00.716- 0.914)/0.0184 |
PP + AP + AP + AP | 0.672 (0.542-0.785)/< 0.0001 | 0.758 (0.692- 0.909)/0.0079 | 0.863 (0.773-0.927)/0.3188 | 0.806 (0.690- 0.893)/0.0475 |
PP + PVP + PVP + PVP | 0.721 (0.584-0.834)/0.0019 | 0.913 (0.807-0.972)/0.1165 | 0.775 (0.675-0.857)/ 0.0247 | 0.900 (0.796- 0.962)/0.7491 |
PP + DP+ DP+ DP | 0.652 (0.513-0.774)/0.0002 | 0.818 (0.692-0.909)/0.0079 | 0.790 (0.688-0.870)/0.0356 | 0.904 (0.802-0.964)/0.7911 |
AP + AP + AP + AP | 0.573 (0.443- 0.696)/< 0.0001 | 0.674 (0.548-0.785)/< 0.0001 | 0.833 (0.739- 0.904)/0.3375 | 0.697 (0.567- 0.807)/0.0019 |
PVP + PVP + PVP+ PVP | 0.697 (0.554-0.817)/0.0029 | 0.859 (0.748- 0.934)/0.0101 | 0.794 (0.693- 0.874)/0.1144 | 0.782 (0.650-0.882)/0.0278 |
DP + DP + DP + DP | 0.697 (0.562- 0.811)/0.0007 | 0.787 (0.666- 0.880)/0.0008 | 0.751 (0.646-0.838)/0.0387 | 0.873 (0.760-0.946)/0.1805 |
- Citation: Cao SE, Zhang LQ, Kuang SC, Shi WQ, Hu B, Xie SD, Chen YN, Liu H, Chen SM, Jiang T, Ye M, Zhang HX, Wang J. Multiphase convolutional dense network for the classification of focal liver lesions on dynamic contrast-enhanced computed tomography. World J Gastroenterol 2020; 26(25): 3660-3672
- URL: https://www.wjgnet.com/1007-9327/full/v26/i25/3660.htm
- DOI: https://dx.doi.org/10.3748/wjg.v26.i25.3660