The Future of Medical Imaging Intelligence Isn’t All Artificial

What does it take to create a new medical-imaging technology-whether hardware or software-that will successfully translate to clinical use and impact patient care? Creativity, innovative thinking, and solid science and engineering? Of course, but that's only one half of the picture. A famous tagline from the 1989 movie Field of Dreams was "If you build it, he will come." That might be true for baseball diamonds, but it's not necessarily the case in medical imaging.

Unfortunately, far too often, technology development in medical imaging fails to consider the ultimate success factor-the human user. Is there even one health care provider that actually likes and looks forward to using their electronic health record? Enough said.

Increasing evidence indicates that radiologists are being negatively impacted by the technology and tools they use in their daily image-interpretation routine. Old-fashioned computer mice, bad chairs, and too much sitting are only partially to blame. The software tools intended to aid radiologists, such as computer-aided detection and decision, may not be having the anticipated impact of interpretation efficacy or efficiency, thus raising questions regarding these tools' utility.

Advances in imaging

Recent special sections in the Journal of Medical Imaging were devoted to advances in breast imaging (forthcoming in 2019) and to the development of artificial intelligence (e.g., deep learning), showcasing techniques for improving image interpretation in a wide variety of imaging applications (spie.org/AIinMI). The proposed technologies and sheer volume of AI-related papers are exciting and will likely change medical imaging-not just radiology but also pathology, telemedicine, and any other facet of medical care that involves images and analyses of complex data streams.

Although the future of AI, AR/VR, and robotics in medicine is bright, for the foreseeable future the physician is still the ultimate decision maker. The technology of today should facilitate and enable their decisions.

Focus on the user

Technological tools with amazing potential risk becoming dust gatherers when they are not well integrated into the users' workflow, or when they present information that adds complexity to the decision process rather than reducing it. Successful enabling tools focus on the user. There are numerous points where the user can-and should-be involved in the technology development process.

Potential users should always be consulted in the very early stages of a technology's development in order to find out what their needs really are. Focus groups are useful for this, but it can be even more effective to embed the tool developer in the ultimate-use environment to observe and record existing protocols and procedures, and to notice the gaps, limitations, or opportunities for improvement. The development of a deep-learning scheme that does something even a first-year resident can accomplish successfully without much effort may not be time well spent, for example. Users should be incorporated throughout development in an iterative fashion to get feedback about the product early and consistently so that changes can be made before it's too late and the resulting product is relegated to the dust bin.

There are a variety of theoretically grounded frameworks for evaluating a tool from the human perspective, but the most influential and commonly used is the Technology Assessment Model (TAM), developed in the late 1980s and refined in many ways since then. The nice thing about TAM is that it not only assesses ease of use, but also usefulness. A tool/technology, no matter how elegantly designed and created, that serves no useful purpose or does not improve the process it is designed to improve, is not worth developing. The tagline for technology, tool, and AI development in medical imaging should perhaps be "If you build it with the user in mind, (s)he will come and win the game!"

Elizabeth Krupinski, PhD, is a professor and vice chair for research at Emory University in the Department of Radiology and Imaging Science. She is President of the Medical Image Perception Society, an SPIE Fellow, and an editorial board member for the Journal of Medical Imaging