From: Artificial intelligence in nonalcoholic fatty liver disease
Study | Country | Study cohort | Diagnostic method | AI classifier | Development cohort (n) | Validation cohort (n) | Validation methods | Sensitivity | Specificity | TP | FP | TN | FN |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Aim: AI-assisted ultrasonography to diagnose NAFLD | NAFLD/total, % Steatosis | NAFLD/total, % Steatosis | |||||||||||
Kuppili et al. | Portugal | Retrospective | Liver biopsy (not defined) | ELMa, SVM | 36/63 | N/A | k-fold cross-validation | 0.913 | 0.921 | 33 | 2 | 25 | 3 |
N/A | |||||||||||||
Byra, et al. | Poland | Prospective | Liver biopsy (> 5% hepatocyte steatosis) | CNN | 38/55 | N/A | 5-fold cross-validation | 1 | 0.882 | 38 | 3 | 15 | 0 |
50% had steatosis < 30% | |||||||||||||
Biswas et al. | Portugal | Retrospective | Liver biopsy (not defined) | CNNa, SVM, ELM | 36/63 | N/A | 10-fold cross-validation | 1 | 1 | 36 | 0 | 27 | 0 |
N/A | |||||||||||||
Shi et al. | China | Prospective | MRI (> 5% hepatic fat content) | RT | 34/60 | N/A | 10-fold cross-validation | 0.875 | 0.9286 | 30 | 2 | 24 | 4 |
92% had steatosis < 20% | |||||||||||||
Han et al. | US | Prospective | MRI (> 5% hepatic fat content) | CNN | 70/102 | 70/102 | Validation cohort | 0.97 | 0.94 | 68 | 2 | 30 | 2 |
Average 11 ± 9% | Average 11 ± 8% | ||||||||||||
Zamanian et al. | Poland | Prospective | Liver biopsy (> 5% hepatocyte steatosis) | CNN + SVM | 38/55 | N/A | 10-fold cross-validation | 0.972 | 1 | 70 | 0 | 74 | 2 |
50% had steatosis < 30% | |||||||||||||
Aim: AI-assisted clinical data sets to diagnose NAFLD | NAFLD/total | NAFLD/total | |||||||||||
Ma et al. | China | Prospective | Ultrasonography | BNa, kNN, SVM, LR, NB, RF, BN, AdaBoost, HNB, Bagging, AODE | 2522/10,508 | N/A | 10-fold cross-validation | 0.675 | 0.878 | 1702 | 974 | 7012 | 820 |
Islam et al | Taiwan | Retrospective | Ultrasonography | LRa, RF, SVM, ANN | 593/994 | N/A | 10-fold cross-validation | 0.741 | 0.649 | 439 | 141 | 260 | 154 |
Wu et al. | Taiwan | Retrospective | Ultrasonography | RFa, LR, ANN, NB | 377/577 | N/A | 10-fold cross-validation | 0.872 | 0.859 | 329 | 28 | 172 | 48 |
Atabaki-Pasdar et al. | United Kingdom | Retrospective | MRI (⩾ 5% hepatic fat content) | RF | 640/1514 | 1011/4617 | Validation cohort | 0.67 | 0.74 | 677 | 838 | 2668 | 334 |
Chen et al | China | Retrospective | Ultrasonography | ANN | Total 10,354 | 2218/4436 | Validation cohort | 0.837 | 0.804 | 1857 | 435 | 1783 | 361 |
Liu et al. | China | Retrospective | Ultrasonography | XGBoosta, LR, SVM, SGD, CNN, MLP, LSTM | 4018/10,373 | 1860/4942 | Validation cohort | 0.611 | 0.909 | 1136 | 280 | 2802 | 724 |
Aim: AI-assisted diagnosis of NASH in patients at-risk for NASH | |||||||||||||
 Gallego-Duran et al | Spain | Prospective | Liver biopsy | LR | NASH/NAFLD | NASH/NAFLD | Validation cohort | 0.87 | 0.6 | 38 | 17 | 26 | 6 |
21/39 | 44/87 | ||||||||||||
 Naganawa et al. | Japan | Retrospective | Liver biopsy | LR | Total 53 | NASH/non-NASH | Validation cohort | No suspicion of fibrosis: 1.00 | No suspicion of fibrosis: 0.92 | 4 | 1 | 11 | 0 |
28-Jul | Suspicion of fibrosis: 1.00 | Suspicion of fibrosis: 0.31 | 3 | 11 | 5 | 0 | |||||||
 Uehara et al. | Japan | Retrospective | Liver biopsy | Rule extraction algorithm | NASH/non-NASH | NASH/non-NASH | Validation cohort | 0.862 | 0.417 | 56 | 7 | 5 | 9 |
79/23 | 65/12 | ||||||||||||
 Garcia-Carretero et al. | Spain | Retrospective | Ultrasonography with LFTs | Lasso regression | NASH/non-NASH | NASH/non-NASH | Validation cohort | 0.7 | 0.79 | 36 | 83 | 314 | 15 |
204/1587 | 51/397 | ||||||||||||
 Docherty et al. | USA | Retrospective | Liver biopsy | kNN, RF, XGBoosta | NASH/NAFLD | NASH/NAFLD | Validation cohort | 0.81 | 0.66 | 146 | 34 | 68 | 34 |
270/152 | 180/102 | ||||||||||||
AI-assisted diagnosis of liver fibrosis in NAFLD | |||||||||||||
 Pournik et al. | Iran | Retrospective | Liver biopsy | ANN | Cirrhotic/non-cirrhotic | Cirrhotic/non-cirrhotic | Validation cohort | 0.657 | 0.987 | 44 | 4 | 309 | 23 |
52/248 | 15/65 | ||||||||||||
 Gallego-Duran et al. | Spain | Prospective | Liver biopsy | LR | F0-1/F2-4 | F0-1/F2-4 | Validation cohort | F2-4 0.77 | F2-4 0.80 | 24 | 11 | 45 | 7 |
20/19 | 56/31 | ||||||||||||
 Shahabi et al. | Iran | Retrospective | Elastography | ANN | F0/F1/F2/F3/F4 | 15% of data set | Validation cohort | F1 0.993 | F1 0.757 | - | - | - | - |
415/151/132/23/5 | (same proportion) | F2 0.939 | F2 0.938 | ||||||||||
F3 1.000 | F3 0.993 | ||||||||||||
F4 1.000 | F4 1.000 | ||||||||||||
 Okanoue et al. | Japan | Retrospective | Liver biopsy and ultrasonography | ANN | Normal/F0/F1/F2/F3/F4 | F0/F1/F2/F3-F4 | Validation cohort | NAFLD (F0) vs. | NAFLD (F0) vs. | 50 | 7 | 1 | 16 |
48/106/74/56/65/23 | 17/18/15/24 | NASH (F1-4) | NASH (F1-4) | ||||||||||
0.877 | 0.941 | ||||||||||||
AI-assisted diagnosis of liver fibrosis in NAFLD | |||||||||||||
 Okanoue et al. | Japan | Retrospective | Liver biopsy | ANN | F0/F1/F2/F3/F4 | F0/F1/F2/F3-F4 | Validation cohort | F0 vs. F1-4: 0.85 | F0 vs. F1-4: 0.867 | 68 | 4 | 26 | 12 |
106/74/56/65/23 | 30/27/24/29 | F0-1 vs. F2-4: 0.755 | F0-1 vs. F2-4: 0.877 | 40 | 7 | 50 | 13 | ||||||
F0-2 vs. F3-4: 0.828 | F0-2 vs. F3-4: 0.877 | 24 | 10 | 71 | 5 | ||||||||
Aim: AI-assisted steatosis quantification of pathological specimen | |||||||||||||
 Vanderbeck et al. | USA | Retrospective | Pathologist | SVM | Macrosteatosis/other features | N/A | 10-fold cross-validation | 0.98 | 0.94 | 1072 | 48 | 859 | 28 |
1100/859 | |||||||||||||
 Liu et al. | China | Prospective | Pathologist | Linear regression | Steatosis grade 0: 0 | Steatosis grade 0: 1 | Validation cohort | Steatosis | Steatosis | 71 | 0 | 1 | 0 |
Grade 1: 77 | Grade 1: 41 | grade 0 vs. ⩾ 1: 0.99 | grade 0 vs. ⩾ 1: 1.00 | 28 | 6 | 36 | 3 | ||||||
Grade 2: 45 | Grade 2: 22 | grade ⩽ 1 vs. ⩾ 2: 0.91 | grade ⩽ 1 vs. ⩾ 2: 0.85 | 6 | 1 | 63 | 3 | ||||||
Grade 3: 24 | Grade 3: 9 | grade ⩽ 2 vs. 3: 0.67 | grade ⩽ 2 vs. 3: 0.98 | ||||||||||
 Sun et al. | USA | Prospective | Pathologist | CNN | 30 | 66 | Validation cohort | ⩾ 30% steatosis | ⩾ 30% steatosis | 15 | 2 | 73 | 6 |
0.714 | 0.973 | ||||||||||||
 Teramoto et al. | Japan | Retrospective | Pathologist | Logistic regression | Matteoni classification46 | Matteoni classification | Validation cohort | type 1 vs. NASH: 0.879 | type 1 vs. NASH: 1.00 | 29 | 0 | 66 | 4 |
Type 1/type 2/type 3-4 (NASH) | Type 1/type 2/type 3-4 (NASH) | type 2 vs. NASH: 0.909 | type 2 vs. NASH: 0.909 | 30 | 6 | 60 | 3 | ||||||
33/33/33 | 33/33/33 |