To see the new trends in radiology in 2018, just take a look back at 2017. AI has come a long way; wearables have become every day, 3D printing has matured, and IoT is, well, meh. Where is Artificial Intelligence (AI) in radiology today? This year, AI is gaining in respect and stature among radiology professionals. Last year in Everything Rad, we reported that AI inspired a mixture of wonder and fear among the radiology community.
Among the difficulties in evaluating AItype medical diagnosis systems are: the intermediate conclusions of the AI system need to be looked at in addition to the "final" answer; the "superhuman human" fallacy must be resisted; both pro-and anti-computer biases during evaluation must be guarded against; and methods for estimating how the approach will scale upwards to larger domains are needed We propose a type of Turing test for the evaluation problem, designed to provide some protection against the problems listed above We propose to measure both the accuracy of diagnosis and the structure of reasoning, the latter with a view to gauging how well the system will scale up A staple of many of the evaluations of AI systems that have so far been conducted (Colby, Hilf, Weber, 81 Kraemer, 1972; Yu et al, 1979) is a central idea from a well-known proposal to evaluate AI systems: The Turing Test (Turing, 1963) The meat of the idea is to see if a neutral observer, given a set of performances on a task, some by a machine and others by humans, but unlabelled as to authorship, could identify, better than chance, which were machine and which were human-produced. Note that this really attempts to answer the question, "DO we know how to design a machine to perform a task which until now required human intelligence?", The latter question subsumes the former in a sense: because the machine not performing well in comparison to a human would presumably increase the cost significantly. In this paper I follow tradition and consider the evaluation of AI systems for medical diagnosis from the viewpoint of the first question above. The proposed procedure is also a variant of Turing's Test.
Eastman Kodak (Rochester, N.Y.), a manufacturer of imaging-related products, has developed an online neural network-based machine vision system for surface mount solder paste inspection. Caere (Los Gatos, Calif.), a provider of neural network-based optical character recognition (OCR) technology, has signed an agreement to supply IBM Ireland with OCR Readers for AN POST, Ireland's national postal service. Using a handheld wand, postal employees will be able to scan text and read bar codes from anywhere on a document. BrainTech (Scottsdale, Ariz.), a developer of neural network and fuzzy logic-based pattern recognition technologies, has signed a development agreement with Raven (Alexandria, Va.), a developer of acoustic systems for the U.S. Navy. BrainTech will integrate its pattern-matching recognition engine into Raven's new medical diagnostic systems.
Artificial intelligence (AI) is booming in all areas of business and medicine. But it's taken a special hold in radiology. RSNA 2017 had its first machine learning showcase, highlighting the companies and technologies already changing how radiology is practiced. About 30 machine learning companies demonstrated their capabilities. AI brings a mix of feelings to those who practice radiology, including fear, hope, and hype, said Woojin Kim, MD, a Chief Information Officer for Nuance's health care division, and an MSK radiologist.
There has been tremendous buzz around Artificial Intelligence in the past year, and I expect it only to increase in 2018. From self-driving cars to computers that create their own languages (and so had to be shut down) - we've heard and read about it all. Some of these stories depict AI accurately but many of them are plain hyperbole. There hasn't been another technological advance in the past that has the power to change our lives so profoundly and is yet so misunderstood. Computer Vision, the branch of AI that deals with making computers process and recognize images better has probably benefited the most from the recent developments in Deep Learning technology.
With nearly three decades of experience as a radiologist, Stephen Chan, M.D., is a wise mentor for younger radiologists coming into the field. But Chan, an associate clinical professor of radiology at Columbia University Medical Center's Harlem Hospital, is also a tech-savvy veteran who embraces AI in radiology, a movement that is sweeping the medical imaging community. At RSNA 2017, medical imaging's signature gathering -- with annual attendance of about 50,000 making it North America's biggest healthcare conference -- Chan moderated an educational panel of young radiologists and radiology residents. The topic was AI in radiology. Of course, AI in radiology is an umbrella that encompasses various forms of machine learning, including deep learning, the AI variant that is probably the most widely used in today's AI-assisted healthcare imaging applications.
The same sense of eager anticipation mixed with abject fear that machine learning has been driving into the heart of radiology is piercing pathology too. F. Perry Wilson, MD, of Yale takes up the technology as poised to transform the latter specialty, which some consider radiology's closest kin, in a lively blog post spurred by a Dec. 12 JAMA study. In the study, Dutch researchers found algorithms to be more accurate than pathologists at detecting lymph node metastases when both were held to a completely objective gold standard. The study "is the best demonstration to date of how machine learning is going to transform medical imaging," writes Wilson, who blogs independently as The Methods Man. After running through the researchers' salient findings, he notes that the study was small, having used slide interpretations from just two sites.
"We're definitely right in the eye of the storm of the hype cycle," Rasu Shrestha, M.D., chief innovation officer at University of Pittsburgh Medical Center, told SearchHealthIT on the busy "technical exhibition," or show, floor. "Having said that, that hype is being driven by an immense amount of hope. Could AI and machine learning solve for the complexities of healthcare?" Langlotz acknowledged that radiology AI has already been through a number of hype-bust cycles in recent decades, but his work and that of colleagues at the Mayo Clinic and The Ohio State University, among others, shows that AI and machine learning have made dramatic progress. Luciano Prevedello, M.D., division chief for medical imaging informatics at The Ohio State University Wexner Medical Center, said at the same deep learning session that "from 2014 to 2015 is when the algorithms started surpassing the human ability to classify" medical image data.
Artificial intelligence (AI) was a key topic at both MEDICA and the RSNA conference this year. But what are its applications in healthcare in general and radiology in particular? And what are the barriers? Dr. Michael Forsting, director of the Institute of Diagnostic and Interventional Radiology and Neuroradiology at Essen University Hospital in Germany talked to MedicalExpo e-magazine about his experiences with AI. MedicalExpo e-magazine: What are the major challenges facing AI in healthcare?