Table 3.
Summary of AI applications for CRC treatment session (nCRT, neoadjuvant radiotherapy; ANN, artificial neural network; AUC, area under curve; KNN, K-nearest neighbors; SVM, support vector machine; NBC, naïve Bayesian classifier; MLR, mixed logistic regression; LR, logistic regression; NN, neural network; BN, Bayesian network; RF, random forest; CPT-11, Irinotecan; IC50, half maximal inhibitory concentration).
| Topic | Task | Dataset | Model | Performance | Year | Ref. |
|---|---|---|---|---|---|---|
| nCRT | nCRT response prediction | Medical records from 282 patients (248 training and 34 validation) | ANN, KNN, SVM, NBC, MLR | ANN model outperformed others: Accuracy: 0.88, AUC: 0.84, Sensitivity: 0.94 |
2020 | [79] |
| nCRT response prediction | 6555 patients’ records from the SEER | LR | 3-year OS rate: 92.4% with pCR; 88.2% without pCR | 2019 | [80] | |
| nCRT response prediction | 98 patients MRI (53 training set and 45 validation set) | SVM, NN, BN, KNN | Test: AUC: 97.8%, Accuracy: 92.8%, Validation: AUC: 95%, Accuracy: 90% |
2019 | [81] | |
| nCRT response prediction | 55 patients MRI | RF | Mean AUC: 0.83 | 2019 | [82] | |
| Chemotherapy | The toxicity of CPT-11 prediction | Demographic data, liver function bloody tests and tumor markers from 20 advanced CRC patients | SVM | Accuracy: 91% for diarrhea, 76% for leukopenia, and 75% for neutropenia |
2019 | [83] |
| Drug IC50 detection | 18,850 organic compounds | KNN, RF, SVM | Accuracy: over 63% | 2018 | [84] |