From: Denouements of machine learning and multimodal diagnostic classification of Alzheimer’s disease
Technique | Classifier used | Modality | Accuracy | References |
---|---|---|---|---|
Pattern recognition | Convolutional neural network and sparse autoencoder | MRI and PET | AD vs NC: 3 × 3 × 3 (90.30%), 5 × 5 × 5 (91.10%), 7 × 7 × 7 (89.80%) MCI vs NC: 3 × 3 × 3 (87.90%), 5 × 5 × 5 (89.10%), 7 × 7 × 7 (89.20%) | [95] |
Pattern recognition | Random forest algorithm | MRI, PET, CSF, and APOE | AD vs NC (91.80%) AD vs MCI vs NC (60.20%) MCI vs NC (79.50%) | [96] |
Pattern recognition | Multi kernel learning | sMRI and DTI | AD vs NC (90.20%) MCI vs NC (79.42%) AD vs MCI (76.63%) | [97] |
Whole-brain parcellation | SVM | MRI and DTI | Multimodal with 73 features (72.40%) Multimodal with 15 univariate features (72.11%) Multimodal with 15 multivariate features (99.60%) | [98] |
Multi-task feature selection | Multi KSVM | MRI and PET | AD vs NC (94.37%) MCI vs NC (78.80%) | [99] |