ML is already being used to screen for the diseases that have been mentioned in OP. Case-in-point: screening in distant villages in Kerala, India; in collaboration with Google. But currently, it may be too early to say that ophthalmologists can be 'replaced' by AI. Ethical issues, lack of widespread validation of ML tools and hesitation in accepting an entirely 'mechanical' physician are some important questions to be tackled before AI becomes the norm.

There exist literature a heap of literature and basic works on diagnosis diseases via "Eyes Images" however., not very serious attention have been given to its commercialization. I think the future of diseases detection/recognition will be mostly based on images analysis.

ML is already being used to screen for the diseases that have been mentioned in OP. Case-in-point: screening in distant villages in Kerala, India; in collaboration with Google. But currently, it may be too early to say that ophthalmologists can be 'replaced' by AI. Ethical issues, lack of widespread validation of ML tools and hesitation in accepting an entirely 'mechanical' physician are some important questions to be tackled before AI becomes the norm.

Assist - definitely yes.

Replace - not yet.

Inconsistencies in data collection between different devices (commercial instruments), such as OCT, fundus image, FA images, should also be noticed.

Find attached an article that addresses your query.

Artificial Intelligence has been used to cost-effectively diagnose diabetic-related eye disease automatically with far more accuracy than current methods…

https://www.drugtargetreview.com/news/38649/ai-eye-disease/

How artificial intelligence is helping to prevent blindness - Nature

A 30-year vision. Abràmoff began to look into automating the detection of eye diseases around 30 years ago.

https://www.nature.com/articles/d41586-019-01111-y

DeepMind's AI can detect over 50 eye diseases as accurately as a ...

Step by step, condition by condition, AI systems are slowly learning to diagnose disease as well as any human ...

https://www.theverge.com/platform/amp/2018/8/13/17670156/deepmind-ai-eye-disease-doctor-moorfields

AI Eye Screening - Eyenuk, Inc. ~ Artificial Intelligence Eye Screening

... the optic nerve head in detail allowing them to make a diagnosis of not only vision threatening eye disorders such as diabetic retinopathy ...

https://www.eyenuk.com/en/ai-eye-screening/

Interesing! Following.

Vision correction specialist - latest technology in eye care - google.com

have your Vision restored with the help of Egyptian ophthalmologist experts. we provide high quality services: Glaucoma-Diabetic- lasik -Squint- cataract- Cosmetic.

https://www.googleadservices.com/pagead/aclk?sa=L&ai=DChcSEwiNzfqIzpTrAhWJ4ncKHdj1BU8YABAAGgJlZg&ohost=www.google.com&cid=CAASEuRooix1uh23z25IWn4Y_i1TPw&sig=AOD64_3XwWnycgEFAshOLOadFeC4rJ9ibA&q&adurl&ved=2ahUKEwjZzfKIzpTrAhWw4YUKHfWpAiUQ0Qx6BAgQEAE

How artificial intelligence is helping to prevent blindness - Nature

... learning is being used to automate the detection of eye diseases. ... Wong likens the rise of AI-based diagnosis to that of ...

.https://www.nature.com/articles/d41586-019-01111-y

Artificial intelligence may help in screening for common ocular morbidities especially in hard to reach terrains but final diagnosis of the disease conditions, treatment initiation and maintenance must be from the ophthalmologist side.

Much more to consider. Our current understanding is evolving. It may replace doctors in the far future but not now.

The current major implication of AI in medicine is image classification (which depends on too many parameters) and image segmentation (say drusen/ hard exudates count). Inconsistencies in data, training models and associated data are some major factors in determining the performance of these systems.

Further, they need large training data for better accuracy and for many diseases (uncommon and rare) we may not have enough data to train AI models.

BUT AI IS EVOLVING RAPIDLY. MAYBE AT SOME POINT WE ALL WILL HAVE REDUCED ROLE IN PATIENT MANAGEMENT OR WILL THEY JUST IMPROVE OUR PERFORMANCE.

the question isn't if AI is ready to replace a human for a limited task which happens to be a healthcare delivery issue here but whether the public at large is ready to accept AI in the current form to determine sight/health-related issues. As the public acceptance improves and the explainability of AI models which are still considered a sort of black boxes improves and with the right training, AI can not only replace an opthalmologist but any specialist for that matter..

The most appropriate application for this type of AI-assistance (at least in the US healthcare system) would be a screening application in primary care. If all ophthalmoscopes used for cursory optic nerve examination could 1. be expanded in terms of field of view captured in the retina, and 2. take a photo, those retinal images could be used by an AI to detect latent disease potentially before any symptoms present. The threshold for detection and segmentation would be likely be lower, at least at first, because the AI would only need to detect disease (yes/no) and not provide specifics or classification. Those patients would then be sent to an ophthalmologist for follow up care.

However, these types of images are very hard to achieve, especially on an undilated eye with a hand held unit. But there is tremendous potential impact...

Hi,

Yes, there are a number of articles in this field that include the following topics:

1. Replication study: Development and validation of deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs

2. Diabetic Retinopathy Detection via Deep Convolutional Networks for Discriminative Localization and Visual Explanation

3. 3D Self-Supervised Methods for Medical Imaging

4. Maximum Likelihood with Bias-Corrected Calibration is Hard-To-Beat at Label Shift Adaptation

5. O-MedAL: Online Active Deep Learning for Medical Image Analysis

6. Deep Learning Approach to Diabetic Retinopathy Detection

7. Transfer Learning based Detection of Diabetic Retinopathy from Small Dataset

I think that in full way - it is not possible. Just like some add feature in diagnostic devices. The human experiences is a much more bigger and precise tool for diagnostic and cure.

Yes, I'm sure this will happen

AI/ ML outperforms when trained on large data sets. As both AMD and DR are quite frequent diseases and it's very easy to have remarkably large data sets to train machines; AI may outperform Human experts in detecting and grading the disease.

At present, we may consider AI as some tool to assist Human experts future development may lead AI to replace us (At least for these two diseases)

No, it is not, but AI certainly is changing rapidly the way we assess and diagnose many ocular diseases through imaging technology.