Table 2.
Summary of representative studies on 18F-FDG PET/CT radiomics-based machine learning analyses in head and neck tumors
| Authors | Years | Tumor type | Aim | Sample size | Constructed ML models | Core ML algorithm | Best ML model | Validation | Resultsa |
|---|---|---|---|---|---|---|---|---|---|
| Differentiating benign from malignant tumors | |||||||||
| Aksu et al. [60] | 2020 | Thyroid incidentaloma | Benign vs. malignant | n = 60 | PET radiomics only | RF | – | Training and validation cohorts | AUC: 0.849 |
| Predicting tumor characteristics | |||||||||
| Haider et al. [62] | 2020 | OPC | HPV status | n = 435 |
Tumor PET/CT Lymph node PET/CT Tumor and lymph node PET/CT |
XGB | Tumor and lymph node PET/CT | Training and validation cohorts | AUC: 0.83 |
| Predicting treatment response or survival | |||||||||
| Haider et al. [63] | 2021 | OPC | Locoregional recurrence after RT | n = 190 |
Clinical model CT radiomics-based model PET radiomics-based model Combined PET and CT model Combined clinical, PET, and CT model |
RSF | Combined model | Internal validation (cross-validation) | C-index: 0.76 |
| Nakajo et al. [64] | 2023 | HPC | PFS after RT, CRT, or surgery | n = 100 | Combined clinical + PET radiomics-based model alone | LR | – | Training and validation cohorts | HR: 3.22 |
| Lafata. et al. [65] | 2021 | OPC | Recurrence-free survival after RT | n = 64 | Intra-treatment PET radiomics-based model | Unsupervised data clustering algorithm | – | Internal validation | HR: 2.69 |
| Spielvogel et al. [66] | 2023 | HNSCC | 3-year OS | n = 127 | Combined genomic + CT radiomics-based + PET radiomics-based model alone | Ensemble ML algorithm | – | Internal validation (cross-validation) | AUC: 0.75 |
| Haider et al. [67] | 2020 | OPC | OS after RT, CRT, or surgery | n = 306 |
Clinical model CT radiomics-based model PET radiomics-based model Combined PET and CT model Combined clinical, PET, and CT model |
RSF | Combined model | Training and validation cohorts |
5-year OS, HPV-associated oropharyngeal cancer (p = 0.02); 5-year OS, HPV-negative oropharyngeal cancer (p = 0.01) |
| Zhong et al. [68] | 2021 | HPC and LC | Disease progression at 1 year after chemotherapy or RT | n = 72 |
CT radiomics-based model PET radiomics-based model Combined model |
RF | Combined model | Training and validation cohorts | AUC: 0.94 |
| Du et al. [69] | 2019 | NPC | Local recurrence after chemotherapy or RT | n = 76 | PET radiomics-based model alone | RF | – | Internal validation (cross-validation) | AUC: 0.892 |
| Peng et al. [70] | 2019 | NPC | 5-year DFS after chemotherapy or CRT | n = 707 | Combined PET radiomics-based + CNN-based model alone | LASSO regression | – | Training and validation cohorts | C-index: 0.722 |
| Liu et al. [71] | 2020 | HNSCC | OS after RT | n = 171 | PET radiomics-based model alone | LASSO regression | – | Internal validation (cross-validation) | C-index: 0.77 |
aPerformance only presents the result of the best machine learning model
AUC area under the receiver operating characteristic curve, C-index concordance index, CNN convolutional neural network, CRT chemoradiotherapy, DFS disease-free survival, HNSCC head and neck squamous cell carcinoma, HPC hypopharyngeal cancer, HPV human papillomavirus, HR hazard ratio, LASSO least absolute shrinkage and selection operator algorithm, LC laryngeal cancer, LR logistic regression, ML machine learning, NPC nasopharyngeal cancer, OPC oropharyngeal cancer, OS overall survival, PFS progression-free survival, RF random forest, RSF random survival forest, RT radiotherapy, XGB gradient tree boosting