Table 5 Characteristics of ML-based predictive model of HCC patients after TACE
Author | Study type | No. of patients | Model | Outcomes | AUC | Conclusion |
|---|---|---|---|---|---|---|
Abajian [77], 2018 | Retrospective Single center | 36 | RF | Responders or non-responders | – | RF model combined with MRI parameters may be predicted tumor response of post-TACE HCC |
Morshid [79], 2019 | Retrospective Single center | 105 | RF | TACE-susceptible or TACE-refractory | 0.733 | The accuracy of RF model using a combination of clinical parameters plus quantitative image features was higher than the RF model based on the clinical parameters alone, in the study of predicting HCC response to TACE |
Mähringer-Kunz [76], 2020 | Retrospective Single center | 282 | ANN | 1-year survival | V: 0.77 ± 0.13 D: 0.83 ± 0.06 | The ANN model had a promising performance at predicting HCC patient survival after TACE and outperformed the traditional scoring systems |
Peng [25], 2020 | Retrospective Multicenter | 798 | CNN | CR, PR, SD, PD | D: 0.97 (CR) 0.96 (PR) 0.95 (SD) 0.96 (PD) V: 0.98 (CR) 0.96 (PR) 0.95 (SD) 0.94 (PD) | The CNN model presented a good performance for predicting the outcome of TACE |
Liu [78], 2020 | Retrospective Single center | 138 | CNN SVM1* SVM2# | ORR | D: 0.98 (CNN) 0.84 (SVM1) 0.82 (SVM2) V: 0.93 (CNN) 0.80 (SVM1) 0.81 (SVM2) | CNN is better in predicting treatment response over SVM in HCC patients treated with TACE |