Table 1 Selected applications of deep learning
From: Opportunities and Challenges: Classification of Skin Disease Based on Deep Learning
Disease type | Diagnostic target | Architecture | Accuracy | Control group |
|---|---|---|---|---|
Skin cancer [9] | Multiple classification | Inception v3 CNN | > 0.960 | ~ 0.880 |
Congenital cataract [35] | Binary classification, object localization | VGG | 0.976‒0.989 | ‒ |
Pneumonia [36] | Multiple classification, object detection | CheXNet | 0.735‒0.937 | 0.633‒0.914 |
Diabetic retinopathy [37] | Binary classification | Inception v3 | 0.840 | 0.800‒0.910 |
Cardiovascular disease [38] | Binary classification | ‒ | 0.764 | 0.728 |
Alzheimer’s disease [40] | Binary classification | LeNet, GoogLeNet | 0.845‒0.988 | ‒ |
Genetic disorders [46] | Binary classification | DeepGestalt | 0.920‒0.969 | 0.710‒0.870 |
Cutaneous tumors[47 | Multiple classification | ResNet-152 | 0.96 | ‒ |
Melanoma[48] | Binary classification | Inception v4 | 0.860 | 0.790 |
Binary classification, logistic regression | CNN | 0.897 | ‒ | |
Cardiac contractile dysfunction [51] | Binary classification | CNN | 0.835‒0.925 | ‒ |
Arrhythmia [52] | Multiple classification | DNN | 0.837 | 0.780 |
Melanoma and Seborrheic keratosis [53] | Binary classification | DCNN | 0.917 | 0.886 |
Skin tumor [54] | Multiple classification | DCNN | 0.924 | 0.853 |
Melanoma [55] | Binary classification | ResNet-152 | 0.944 | 0.823 |