Table 4.
Comparison of Performance Between the Proposed Model and From Other Studies.
| Author | Detections | Algorithms | AUC | Sensitivity | Specificity |
|---|---|---|---|---|---|
| Long 25 | Multi-platform transcriptomics | RF | 0.998 (0.995-0.999) | 99.8% | 99.9% |
| Nakajima 20 | Urinary polyamine biomarker panel | ADTree | 0.961 (0.937-0.984) | N/A | N/A |
| Wan 22 | Whole-genome sequencing | LR + SVM | 0.92 (0.91 to 0.93) | 85% | 85% |
| Hornbrook 19 | Complete blood count | ColonFlag® | 0.80 (0.79-0.81) | N/A | N/A |
| Kinar 28 | Complete blood counts | GBM + RF | 0.82 | 50% | 87% |
| Zhao 29 | Age, BMI, gut bacteria | LR + SVM | 0.942 | 93.3% | 80.7% |
| Proposed | CEA, hemoglobin, HDL, and Lp(a) | LR | 0.849 (0.840-0.860) | 88.3% | 81.5% |
Abbreviations: ADTree: alternating decision tree; AUC: area under the curve; BP: backpropagation; CEA: carcinoembryonic antigen; GBM: gradient boosting model; LR, logistic regression; RF, random forests; SVM, support vector machine.