Table 7.
Applications of artificial intelligence in biomarker-based pancreatic cancer diagnosis
| Reference | Sample Size | Data Source | Algorithm | Aim | Best result |
|---|---|---|---|---|---|
| Chen et al. 173 | 28 samples* | DNA-PAINT (exosomes) | LDA | Cancer detection | Accuracy (100%) |
| Zheng et al. 174 | 220 cases** | MALDI-TOF-MS (exosomes) | ANN | Cancer discrimination | AUC (0.86) |
| Ko et al. 175 | 28 mice + 34 cases | ExoTENPO chip (exosomes) | LDA | PC diagnosis | Accuracy (100%) |
| Gao et al. 176 | 199 cases | SELDI-TOF-MS (proteomes) | SVM, KNN, ANN | PC diagnosis | AUC (0.971), Sensitivity (96.67%), Specificity (100%) |
| Yu et al. 177 | 100 serum samples | SELDI-proteinchip | DT | PC prediction | Sensitivity (88.9%), Specificity (74.1%) |
| Yang et al. 178 | 913 serum samples | Multiple serum tumor markers | ANN, LR | PC diagnosis | AUC (0.905), Accuracy (83.53%), Sensitivity (90.86%), Specificity (67.50%) |
| Qiao et al. 179 | 136 cases | CT images+ serum tumor markers | 2D-3D CNN | Image segmentation; PC vs CP |
For image segmentation: DSC (84.32%); For PC vs CP: Accuracy (87.63%), Sensitivity (94.57%), Specificity (93.25%), PPV (84.57%), NPV (90.34%) |
| Cristiano et al. 77 | 34 cases | Cell-free DNA | GBM | Cancer detection | AUC (0.86), Accuracy (67%), Specificity (71%) |
| Alizadeh Savareh et al. 180 | 671 cases | GEO database (circulating microRNA) | PSO-ANN-NCA | PC diagnosis | Accuracy (93%), Sensitivity (93%), Specificity (92%) |
| Yu et al. 181 | 501 cases | exLR | SVM | PDAC detection | AUC (0.960), Accuracy (90.43%), Sensitivity (93.39%), Specificity (85.07%) |
| Almeida et al. 182 | 648 samples | Gene expression microarray | ANN | PDAC prediction | F1-score (0.86), Accuracy (89.66%), Sensitivity (87.6%), Specificity (83.1%) |
| Yang et al. 197 | 204 cases | Liquid biopsy | KNN, SVM, LDA, LR, and Naive Bayes | PC diagnosis and staging |
For diagnosis: AUC (0.95), Accuracy (92%), Sensitivity (88%), Specificity (95%); For staging: Accuracy (84%), Sensitivity (78%), Specificity (88%) |
| Sinkala et al. 198 | 185 cases | TCGA database (proteins, mRNAs, miRNAs, and DNA methylation patterns) | NCA, SVM, DT, LR, ET, KNN | PC subtypes differentiation | Accuracy (98.7% for mRNA-based KNN classifier; 97.8% for the DNA methylation-based SVM classifier) |
| Zhang et al. 199 | 1183 cases*** | LDI-MS | SVM | Pan-cancer diagnosis and classification | For PC: Accuracy (100%) |
*Including 9 healthy samples, 10 breast cancer samples, 9 PC samples;
**Including 79 breast cancer cases, 57 PC cases, 84 healthy controls;
***Including 97 PC cases.
Abbreviations: ANN: artificial neural network; AUC: area under the curve; CNN: convolutional neural network; CP: chronic pancreatitis; DT: decision tree; DNA-PAINT: DNA points accumulation for imaging in nanoscale topography; ET: ensemble tree; exLR: extracellular vesicles long RNA; GBM: gradient tree boosting; KNN: k-nearest neighbor; LDA: liner discriminate analysis; LDI-MS: laser desorption/ionization mass spectrometry; LR: logistic regression; MALDI-TOF-MS: matrix-assisted laser desorption/ionization time-of-flight MS; MLP: multilayer perceptron; NCA: neighborhood component analysis; PC: pancreatic cancer; PDAC: pancreatic ductal adenocarcinoma; PPV: positive predict value; SELDI-TOF-MS: surface-enhanced laser desorption/ionization time-offlight mass spectrometry; SVM: support vector machine.