Table 4 Comparison of performance metrics of MLF I and MLF II with other state of the art classification models.
From: Radiomics based likelihood functions for cancer diagnosis
| Â | Prediction Model | # of Data sets (Tumor site, Database) | Accuracy | Sensitivity | Specificity | AUC |
|---|---|---|---|---|---|---|
Wu et al.8 | Random forest classifier | 152(Lung, LIDC) | 55.0% | 80.0% | 72.0% | — |
Chen et al.9 | SFS, SVM | 75(Lung, LIDC) | 84.0% | 92.85% | 72.73% | — |
Choi et al.10 | SVM-LASSO | 72(Lung, LIDC) | 84.6% | 87.2% | 81.2% | 89% |
Liu et al.11 | Multi-view convolutional neural networks | 172(Lung, LIDC) | 94.59% | — | — | 98.1% |
Kumar et al.13 | Deep convolutional neural network | 97(Lung, LIDC) | 75.1% | 83.35% | 61.0% | — |
Pallamar et al.37 | Linear Discriminant analysis, k nearest neighbor | 27(Head & Neck, Private) | 81.48% 1.5T 92.59% 3T | — | — | — |
Huang et al.38 | Gene expression | 462(Colon, Private) | — | — | — | — |
Proposed MLF I | Curve fitting using non-linear regression | 200(Lung, LIDC & Lung1) 35(Colon, CTC) 30(Head & Neck, HNSCC) | 91.5% 74.28% 83.33% | 95.68% | 73.68% | 92.68% |
Proposed MLF II | Curve fitting using non-linear regression | 200(Lung, LIDC & Lung1) 35(Colon, CTC) 30(Head & Neck, HNSCC) | 97% 85.71% 90% | 98.77% | 89.19% | 98.81% |
