Table 2.

Applications of explainability in medical imaging.

Method	Algorithm	Model	Application	Modality
Attribution	Gradient*I/P, GBP, LRP, occlusion [36]	3D CNN	Alzheimer’s detection	Brain MRI
	GradCAM, GBP [37]	Custom CNN	Grading brain tumor	Brain MRI
	IG [38]	Inception-v4	DR grading	Fundus images
	EG [39]	Custom CNN	Lesion segmentation for AMD	Retinal OCT
	IG, SmoothGrad [41]	AlexNet	Estrogen receptor status	Breast MRI
	Saliency maps [42]	AlexNet	Breast mass classification	Breast MRI
	GradCAM, SHAP [49]	Inception	Melanoma detection	Skin images
	Activation maps [50]	Custom CNN	Lesion classification	Skin images
	DeepDreams [46]	Custom CNN	Segmentation of tumor from liver	CT imaging
	GSInquire, GBP, activation maps [47]	COVIDNet CNN	COVID-19 detection	X-ray images
Attention	Mapping between image to reports [56]	CNN & LSTM	Bladder cancer	Tissue images
	U-Net with shape attention stream [57]	U-net based	Cardiac volume estimation	Cardiac MRI
Concept vectors	TCAV [59]	Inception	DR detection	Fundus images
	TCAV with RCV [60]	ResNet101	Breast tumor detection	Breast lymph node images
	UBS [61]	SqueezeNet	Breast mass classification	Mammography images
Expert knowledge	Domain constraints [63]	U-net	Brain MLS estimation	Brain MRI
	Rule-based segmentation, perturbation [64]	VGG16	Lung nodule segmentation	Lung CT
Similar images	GMM and atlas [6]	3D CNN	MRI classification	3D MNIST, Brain MRI
	Triplet loss, kNN [65]	AlexNet based with shared weights	Melanoma	Dermoscopy images
	Monotonic constraints [66]	DNN with two streams	Melanoma detection	Dermoscopy images
Textual justification	LSTM, visual word constraint [67]	Breast mass classification	CNN	Mammography images
Intrinsic explainability	Deep Hierarchical Generative Models [68]	Auto-encoders	Classification and segmentation for Alzheimer’s	Brain MRI
	SVM margin [69]	Hybrid of CNN & SVM	ASD detection	Brain fMRI