Table 1.

Analysis of various methods in ADHD detection using CNNs.

Ref. No	Insights	Methods Used	Results	Limitations
(16)	Effective ADHD diagnosis using EEG signals; importance of specific brain regions and frequency bands identified	EEG signals, preprocessing, frequency feature extraction, CNN, Layer-wise Relevance Propagation	94.52% accuracy	Focused only on EEG data; small sample size
(17)	Enhanced diagnostic accuracy using FIVLNet framework combining MRI image data	FIVLNet, high and low-frequency data, CNN, cross-attention mechanism, textual embeddings (CLIP)	93.89% accuracy on fMRI data	Focuses on MRI data; does not address multimodal data
(18)	Differentiated adults with ADHD from healthy individuals using event-related spectral EEG data	Event-related spectral EEG data, CNN	88% classification accuracy	Limited to spectral EEG data; specific to adult population
(19)	ECG-based DL model for ADHD detection	One-dimensional CNN, dropout, ReLU activation, L2 kernel regularization, Adam optimizer	Significant feature reduction, accurate classification	Limited to ECG data; potential overfitting not extensively tested
(20)	High accuracy in ADHD diagnosis using fMRI data with different models	fMRI data, Nadam, SGDM, proposed CNN	Proposed CNN: 98.77% accuracy	Specific to fMRI data; does not explore multimodal data
(21)	Developed 3D CNN for classifying ADHD using MRI scans	3D CNN, multimodality architecture combining fMRI and sMRI data	69.15% accuracy	Lower accuracy compared to other methods
(22)	Diagnosed ADHD using deep multimodal 3D CNN with gray matter and fALFF features	ADHD-200 dataset, deep multimodal 3D CNN, KNN, SVM, LDA	LDA: 74.93% accuracy	Only uses ADHD-200 dataset; compares limited classifiers
(23)	Combined CDAE and RF for superior ADHD classification	Convolutional Denoising Autoencoders, Random Forest, ensemble learning, grid search optimization	75.64% accuracy, 76.922% sensitivity, 73.08% specificity	Needs larger datasets; explores limited feature extraction techniques
(24)	Precise brain tumor classification using RBP and GSN preprocessing	RBP, Gray Standard Normalization (GSN), CNN	96% accuracy, 7% false classification rate on a dataset of 3000 samples	Specific to brain tumor classification; does not address ADHD
(25)	Comprehensive review and meta-analysis of sMRI and fMRI-based ML techniques for ADHD diagnosis	sMRI, fMRI, ML techniques, sensitivity, specificity, likelihood ratios, AUC, meta-regression	Sensitivity: 0.74, Specificity: 0.75, AUC: 0.81	Excludes EEG-based methods; needs further analyses on multimodal data
(26)	Precise brain tumor classification using RBP and GSN preprocessing	RBP, Gray Standard Normalization (GSN), CNN	96% accuracy, 7% false classification rate on a dataset of 3000 samples	Specific to brain tumor classification; does not address ADHD
(27)	Comprehensive review and meta-analysis of sMRI and fMRI-based ML techniques for ADHD diagnosis	sMRI, fMRI, ML techniques, sensitivity, specificity, likelihood ratios, AUC, meta-regression	Sensitivity: 0.74, Specificity: 0.75, AUC: 0.81	Excludes EEG-based methods; needs further analyses on multimodal data