Table 2.
Overview of study aim, ML method, and performance for the included studies
| Authors | Study aim | ML methodology | Performance |
|---|---|---|---|
| AlKubeyyer et al. 2020 [29] | Development of a computer-aided detection of the meningioma tumor firmness |
• Support vector machine • k-nearest neighbor |
• F-score=0.95 • Balanced accuracy= 0.87 • AUC=0.87 |
| Arokia Jesu Prabhu et al. 2018 [30] | Automatic classification of parasagittal meningioma | Support vector machine | Accuracy= 0.92 |
| Chen et al. 2019 [31] | Automatic classification of meningiomas |
• Linear discriminant analysis • Support vector machine |
Accuracy=0.76 |
| Chu et al. 2020 [11] | Prediction of meningiomas grade | Logistic regression |
• Accuracy= 0.95 (training group) and 0.93 (test group) • Sensitivity= 0.94 training group) and 0.92 (test group) |
| Florez et al. 2018 [32] | Differentiation of vasogenic from tumor cell infiltration edema for radiotherapy | Linear regression | AUC>0.71 |
| Hamerla et al. 2019 [33] | Differentiation of low grade from high grade meningioma |
• Random forest • Extreme gradient boosting • Support vector machine • Multilayer perceptron |
AUC= 0.97 (Extreme gradient boosting) |
| Kanazawa et al. 2018 [34] | Distinction of solitary fibrous tumor/hemangiopericytoma from angiomatous meningioma | Texture analysis |
• Positive predictive value=0.63 • Specificity=0.63 |
| Ke et al. 2019 [35] | Differentiation between benign and non-benign meningiomas | • Support vector machine |
• AUC= 0.91 • Accuracy= 0.89 • Sensibility=0.93 • Specificity=0.87 |
| Laukamp et al. 2018 [36] | Automatic detection and segmentation of meningioma | Deep learning |
• Detection accuracy=0.98 • Mean Dice coefficient for total tumor volume =0.81 ± 0.10 |
| Laukamp et al. 2019 [37] | Prediction of meningioma grade | Multivariate logistic regression model | AUC=0.91 |
| Li et al. 2019 [38] | Automatic differentiation of malignant hemangiopericytoma from angiomatous meningioma | Texture analysis | AUC=0.90 |
| Lu et al. 2018 [39] | Prediction of meningioma grade using ADC maps |
• Classic decision tree • Conditional inference • Decision forest |
Accuracy= 0.62 |
| Morin et al. 2019 [40] | Prediction of meningioma grade, local failure and overall survival | Random forest |
• Grade= Accuracy 0.65; AUC 0.71 • Local Failure= Accuracy 0.61, AUC=0.68 • Overall Survival= accuracy 0.67, AUC= 0.75 |
| Niu et al. 2019 [41] | Differentiation of meningioma subtypes | Fisher discriminant analysis | Accuracy= 0.99-0.1 |
| Park et al. 2018 [42] | Prediction of grade and histological subtype |
• Support vector machine • Random forest |
AUC= 0.86 |
| Speckter et al. 2018 [13] | Prediction of response after radiosurgery | Texture analysis | Correlation coefficient=−0.64 |
| Tian et al. 2020 [43] | Contrastive analysis between craniopharyngioma and meningioma | Binary logistic regression | AUC>0.70 |
| Wei et al. 2020 [44] | Differentiation of hemangiopericytoma from meningioma | Logistic regression model | AUC= 0.92–0.99 |
| Yan et al. 2017 [45] | Prediction of meningioma grade |
• Logistic regression • Naïve Bayes • Support vector machine |
• AUC= 0.73–0.88 • Sensitivity= 0.48–0.91 • Specificity= 0.70–0.96 |
| Zhang et al. 2019 [12] | Prediction of recurrence in skull base meningiomas | Random forest | Accuracy= 0.90 |
| Zhang et al. 2020 [46] | Discrimination of lesions located in the anterior skull base |
• Linear discriminant analysis • Support vector machine • Random forest • Adaboos • K-nearest neighbor • GaussianNB • Logistic regression • gradient • boosting decision tree • Decision tree |
AUC>0.80 |
| Zhu et al. 2019 [47] | Automatic prediction of meningioma grade | Convolutional neural network | AUC= 0.83 |
| Zhu et al. 2019 [48] | Automatic prediction of meningioma grade | Deep learning |
• AUC= 0.81 • Sensitivity= 0.8 • Specificity=0.9 |
AUC area under the receiver operating characteristic curve