Table 14.
Research Gap & Issue | Description | Solution |
---|---|---|
Clinical Results | Ophthalmologists feedback is required in order to check the accuracy of the deep learning predictor. |
Cross-Database Validation |
Data Augmentation | Accurate data augmentation is an expensive solution and expert Ophthalmologist services are required in every angle of lesion image. |
Generative Adversarial Networks (GANs), New data augmentation techniques with fewer learnable parameters |
Class Imbalance | The number of DR cases is much lower than normal cases |
Data Augmentation Techniques, Geometric Transformations |
Lack of Uniformity | Angle of images are not uniform, out of focus, and causes the diffusion of light in the retina |
Generative Adversarial Networks (GAN) New Augmentation Techniques |
Translation Effect | Variability and screening programs do not follow a standard and cause issues |
Translation Standards are required |
Race Scaling | It has been observed that darker retina vascular properties are comparatively different to the light tone retina |
Heterogeneous cohorts parameters |