A team of international researchers belonging to the United States, China, and Australia has invented a new system of AI. They developed a comprehensive solution of a retina specialist system based on AI, called the Care System. This system can identify and monitor a wide range of retinal diseases.
The researchers are from the University of Miami Miller School, Monash University, Capital Medical University, Beijing Tongren Eye Center, Sun Yat-sen University, and Beijing Eagle Vision Technology. The Care System utilizes fundus photography combined with an in-depth learning system extensively trained to detect retinal disease.
Researchers said that they developed this Care System using fundus photography in conjunction with the deep learning system. This deep learning system involves training using data from real-world retinal disease cases. Next, they externally tested this system with the help of fundus photographs already collected from clinical settings the model would most possibly be utilized.
Fundus photography is the process of shooting serial snapshots of the inner part of the eye to screen for retinal abnormalities through the pupil. A fundus camera comprises a low-power specialized microscope and a camera used to observe eye structures. These structures include the lens, retina, and optic disc.
Zongyuan Ge is one of the engineering associate professors at Monash University in electrical and computer systems. He said that researchers accomplished the Care System to identify the 14 widely diagnosed retinal diseases.
For this purpose, they used 207,228 different colors of fundus photos obtained from 16 clinical settings worldwide. These regions include North America, Africa, Asia, and Europe, covering a variety of retina diseases.
As far as internal validation of the Care System is concerned, they used 21,867 photographs to validate the procedure. On the other hand, they externally tested it by using 18,136 photographs. Additionally, they prospectively collected all these photographs from 35 real-world examples across different settings of China. These include 21 physical examination centers, six community hospitals, and eight tertiary hospitals.
The researchers further compared the performance of the Care System with that of 16 ophthalmologists. Whereas they tested it using new camera types and datasets of non-Chinese origin. Based on the outcomes of these tests, Dr. G said that there was a similarity between the performance of the Care System and the professional ophthalmologist when they were tested using non-Chinese datasets. However, the system maintained strong identification performance.
He added that the study's outcomes show that the system is more accurate than a professional. Therefore, the findings allow researchers for further testing on a large scale.
Researchers hope that Care System will be commercially available in China and then in the Asia-Pacific region. They also planned to create a database that would have the screening images taken from real-world environments. Hopefully, this study will improve retinal abnormalities in clinical settings and enhance the diagnosis of retinal diseases.
Furthermore, Amitha Domalpally is a director at the University of Wisconsin in the Madison Imaging Diagnostic Center. She said that through this work, people could expect to see more technological advances in this interplanetary.
Artificial intelligence (AI) is the computer or robot's capability to do usual human tasks. As these tasks require human intelligence and judgment, it is called artificial intelligence.
There are four types of artificial intelligence, including limited memory, reactive machines, theory of mind, and self-awareness.
1. Reactive machines
It is the most basic type of artificial intelligence that is virtuously reactive. This type of AI system neither needs to form memories nor use past practices to respond to recent decisions. Deep Blue is the best example of this type of AI. It is a chess-playing supercomputer of IBM that beat Garry Kasparov, an international grandmaster, in the late 1990s. Google's AlphaGo is the other example of this type of AI.
2. Limited Memory
The limited memory type of artificial intelligence can gather information and store previous data and predictions while considering possible decisions. They essentially look at past experiences to see what might happen next. Moreover, it is more complex than reactive machines, thus offering more possibilities. Three models are using limited memory AI. They include reinforcement learning, Long Short Term Memory (LSTM), and Evolutionary Generative Adversarial Networks (E-GAN).
3. Theory of the mind
The concept of the mind is only ideological. However, we have not acquired the scientific and technical skills yet necessary to meet this level of artificial intelligence.
The idea is wholly based on psychological evidence to understand other living organisms' thoughts and emotions. These thoughts and feelings primarily affect the behavior of someone. Based on AI machines, it means that AI could understand what animals, humans, and other machines feel and how they make decisions through self-determination and self-reflection. Next, they use that information to make their own decisions.
Once research establishes a theory of mind in artificial intelligence, sometimes in the future, the final step would be to become self-aware of AI. This type of AI has a consciousness of the human level. Furthermore, it understands its existence in the world and knows the other's emotional state and presence. Therefore, they could understand the needs of others based on what they communicate with them and how they communicate with them.
A few examples of Narrow AI include:
1. Google search
2. Self-driving cars
3. Image recognition software
4. IBM's Watson
5. Alexa, Siri, and other personal assistants
7. Conversational bots
8. Netflix's recommendations
9. Email spam filters
Retinal diseases are the damages of any part of the retina. Untreated retinal diseases can cause severe loss of vision and even blindness. On the other hand, you can treat some retinal diseases through early detection. At the same time, you can also slow down or control to preserve or even restore the vision in untreatable retina diseases.
Some of the widely observed signs and symptoms of retinal diseases are as follows:
1. Flashes of light or seeing floaters
2. Distorted or blurred vision
3. Reduced vision
4. Blind spots in the central vision
Some factors may involve increasing the risk of retinal diseases, such as:
4. Eye trauma
6. Family history
Retinal detachment is the leading cause of retinal disease. It usually occurs when any fluid comes across a retinal tear. It causes detachment of the retina from the other tissues located at the back of the eye. As a result, it cut off the blood supply to the retina. Therefore, the retina cannot function properly.
Annually, retinal detachments affect around every fifth person out of 100,000 people worldwide. Among them, most cases belong to the elderly and middle-aged groups. Annually, approximately 20 elders and middle-aged adults in 100,000 people were diagnosed with retinal detachment.
There are three types of retinal detachments:
It is the most prevalent type of retinal detachment caused due to the retinal tear. It may be caused by an eye injury, surgery, aging, or nearsightedness.
It usually occurs due to diabetes, when blood vessels placed at the back of the eye are damaged for some reason. It forms scar tissue that pulls on the retina and causes the separation of the retinal tissue from the retina.
It occurs due to building up fluid behind the retina. It leads to pushing the retina away from the retinal tissue attached behind the eyes. Exudative is usually the result of injury, age-related macular degeneration, or inflammation of the retina
An international team of researchers from the United States, China, and Australia has developed a new system of AI, the Care System. They consider this AI system a comprehensive solution for retinal disease, as it can identify and monitor a wide range of retinal diseases.
The Care System works using Fundus photography combined with an extensively trained in-depth learning system to detect retinal disease.
Researchers used a database containing real-world examples of retinal disease cases. The Care System is externally and internally validated using many fundus photographs from various medical settings. Furthermore, the performance of this system is better than the performance of professional ophthalmologists.