Artificial intelligence has been a fascinating concept for years now. Although the technology has grown immensely over the last decade in various fields, including machine learning and also in narrow artificial intelligence, researchers still wonder if general or high-level artificial intelligence can be a reality, especially in radiology.
There is a lot of speculation about artificial intelligence in radiology, and computers replacing radiologists. Eliot Siegel, MD addresses the hype and myth created around artificial intelligence in radiology in his webinar titled “Machine Learning and Artificial Intelligence: Hype, Myth Reality and How It Will Revolutionize the Practice of Quality Diagnostic Imaging”
Role of Artificial Intelligence in radiology in the near future?:
To understand how much of a role AI has to play in the near future, let's understand what AI is. AI is defined as the ability of a computer program or a machine to think and learn. It is also a field of study which tries to make computers "smart". In other words, it is the ability of a computer system to perform tasks that normally require human intelligence, such as speech recognition, decision-making, etc. Siri, Alexa, and Google are some of the famous AI applications to name a few.
What is the scope of AI in radiology?
Mr. Siegal in his webinar says that the artificial intelligence that we have right now is just a narrow version of AI. Narrow Al can help identify or address one particular issue but still cannot perform tasks as precisely as a human. “It specializes in one area. So, you could go online and download a software program for free that would run practically on any computer that could beat the world chess champion Mag at chess, but that same program can’t beat a 5-year-old at tic-tac-toe because it is designed specifically for chess.” Says Mr. Siegal. To explain this he uses a picture from a famous Highlights Magazine.
What’s wrong with the picture?
As we can all see certain objects in the image are inappropriately placed. If you show this to a 5-year-old he might be able to identify what is wrong with the picture. The current AI can identify the objects in the picture, but can't identify what is wrong.
How does this picture justify the scope of AI in radiology? The current artificial intelligence may identify an organ in an MRI scan with immense accuracy, but the problem in radiology is not identifying the organ but figuring out what the flaw is, which takes years of practice for a radiologist himself. Can the current artificial intelligence fulfill this? To attain this precise level of decision making we need artificial general intelligence. The artificial general intelligence is as smart as a human and could replace a human completely.
Where are we with artificial general intelligence?
According to an artificial general intelligence conference, we are nowhere near the human level AI, and a survey conducted in the same conference seems to support the same fact. 42% say we can attain this by 2030, 24% say by 2050, 20% by 2100, 10% after 2100 and 2 % of them say we will never be able to achieve human-level artificial intelligence.
What is your take on artificial intelligence in radiology? Let us know in the comments below.
Also, stay tuned to our next post to know about Machine learning in radiology.