What is artificial intelligence (AI) and how is it being used in Radiology? Artificial intelligence is just a computer system that can mimic human intelligence (5). Think of all the smartphones that have online assistants like, Siri or Bixby; they are AI (5). Technology has had many advances throughout the years in our day to day lives, so why not make medical advances with technology. The field of diagnostic imaging is already introduced to artificial intelligence with computer-aided diagnosis (2). The increasing talk of AI in radiology was spearheaded by Silicon Valley, which lead to discussions at the most prestigious radiology platforms about AI (5). When it comes to diagnostic imaging, patients come in many different habitus. Since AI is just a computer filled with algorithms to only mimic the human mind; it is restricted to the template that the algorithm gives (4). This restriction can cause misdiagnosis or lack of.
In radiography, the computed and digital systems both use computer systems to help improve the quality of the images produced. Technologists can use the computer where the images come up and edit the exams by window-leveling, annotating, flipping the images to present them correctly, etc. This post-processing is used to try to improve the image for the radiologist. This shows how familiar technologists are with using computers. AI is more used on the radiologist side though. Radiologists use Computer-aided diagnosis (CAD) which contains tools that may include image processing, image feature analysis and data classification (5). This CAD tool is used mostly with mammography, and can alert the radiologist to any possible pathologies, like breast cancer, that it ‘sees’ (5).
Hospitals rely heavily on medical imaging to help diagnose their patients. It would be useful if AI could confidently filter out normal images and flag abnormal images for a radiologist to look over (5). In the beginning stages of implementing an artificial intelligence algorithm could be limited to a tool that would have to be activated by the user (5). Though, after becoming comfortable working with AI, it may be granted freedom to report simple scans (5). The algorithm used as diagnostic AI can become even more precise over time after it is given feedback from the simpler exams studied (1).
Since AI could become so great at diagnosing, some might fear that radiologists might become a thing in the past, but there is more to a radiologist’s job than just reading diagnostic images (3). Some other job responsibilities of a radiologist includes:consulting with other physicians on diagnosis and treatment, perform image-guided medical interventions (interventional radiology), relate findings from images to other medical records and test results, and discuss procedures and results with patients (3). If AI did advance to the point it could completely diagnose diagnostic images by itself; the radiologists would have more time to interact with their patients and can focus on doing the other aspects of their jobs.
Overall, AI is not going to take over the radiology world overnight. Even to just be used as a second-hand diagnostic tool, will take years of medical data to be input before the algorithm can fully diagnose alone. Patients and their bodies are so different, which is why we cannot just provide broad spectrum stencils for the algorithm to work from. Artificial intelligence can be an extremely useful tool in the radiology field, not just for the healthcare professionals but for their patients as well.