Table 1.

DL techniques for feature selection on neuroimaging data.

Reference	DL Technique	Accuracy
[69]	SAEsoftmax” regression layer	>86%
[70]	3D-CNN	>87%
[72]	SAE SoftMax” regression layer	>90%
[73]	SAE DNN	>84% for AD/CN classification >82% for MCI to AD classification
[74]	3D CNN	>92% for AD/CN classification >72% for MCI to AD conversion
[75]	VoxCNN ResNet	>79%
[77]	2D CNN	>85%
[78]	3D CNN	>75% for MCI to AD conversion
[80]	SAE 3D CNN	>90%
[81]	Ensemble of 2D CNN and RNN	>91%
[83]	3D CNN	>95% for AD/CN classification >84% for MCI to AD conversion