Radiology is a medical specialty that uses imaging to diagnose and treat diseases in the body. The field of radiology was born in 1895 with the discovery of the x-ray by German physicist and future Nobel Prize winner Wilhelm Röntgen. Since then, radiology has continuously evolved into the field it is today with new technologies (CT, MRI, mammography, fluoroscopy, ultrasound, nuclear medicine) and advancements constantly being developed.
Radiology has become an integral part of medical care, aiding in the diagnosis of innumerable diseases and helping guide disease management. Radiologists, the physicians that interpret medical imaging exams, are the (typically) invisible members of a healthcare team, operating behind the scenes.
We are witnessing an incredible technological revolution affecting all facets of life and have already seen tremendous gains in medicine. So where does it go from here?
In this article, I’ll be discussing my thoughts and predictions on the future of radiology. Will it be rainbows and unicorns or Skynet? Read on!
Radiology is a diverse medical specialty with significant variation between and within radiology practices. So let’s break this down into a few core areas and how they may evolve over time.
Diagnostic imaging, also known as medical imaging, is the skeleton of radiology (pun intended). It is the foundation of every radiology practice as all radiologists interpret imaging exams.
Diagnostic imaging consists of the creation or obtainment of medical images for a clinician, radiologists and referring clinicians alike, to review.
Radiology volumes have increased dramatically over the past few decades with the gradual clinical adoption of radiology modalities over this time. Medical imaging has helped transform medicine and now plays an intricate role in patient care that doctors and patients depend on.
Diagnostic imaging machines consist of two key components (like many machines) – hardware and software.
The machines themselves, the actual scanners, are the hardware. And medicine has seen substantial improvements of scanners across all imaging modalities. Detectors get smaller, heat dissipation techniques become more efficient, MRI gradient coils get stronger, etc.
This is a trend we’re likely to continue to see, though advances in hardware will likely be outpaced by advances in software.
Software advances are where the money really is. While engineers will continue to likely find modest improvements to existing hardware, they will eventually reach an upper limit of what is physically possible. Hardware has likely reached ‘middle age’ whereas software is entering an adolescent growth spurt.
X-ray and CT have actually been using a basic form of artificial intelligence (AI) for a long time in the form of automated dose rate control or modulation. A tiny amount of “test” radiation is used prior to acquiring the image to determine exactly how much radiation is required to create an image of diagnostic quality. Better put, this determines exactly how little radiation can be used to create a quality image. Automated dose control in x-ray and CT has resulted in substantial dose savings since their implementation.
Another big area of interest and research involves how images are processed. CT imaging shifted from filtered back projection to iterative reconstruction decades ago resulting in impressive dose reductions for a standard CT. The software got “smarter” and required less data (radiation) to create similar images.
Model-based iterative reconstruction (MBIR) is a newer technique that is very computational intensive (and therefore time consuming) and can create images from very low doses that would be unreadable in their noisy baseline state. As computational speed continues to increase, MBIR could become the new gold standard in post-processing and result in continued improvements in dose reduction.
Advances in MRI post-processing are currently in development and are poised to reduce image acquisition time. This should allow for faster scan times and decreased artifact from patient movement thanks to the faster scan times. These software algorithms will require robust validation to ensure that clinical care is not negatively impacted.
New interventional radiology devices are continuously being created and new procedures discovered. Minimally invasive procedures and treatments will continue to grow and have a positive impact on patient care, recovery times, and patient outcomes.
Development of new procedures and advancements of existing procedures will continue to increase the positive impact of interventional radiology. There will be a continued shift toward minimally invasive procedures, which will continue to improve patient care, decrease hospital lengths of stay, decrease systemic side effects with directed cancer treatments, and decrease costs to the healthcare system.
Radiology clinical practice, similar to other fields of medicine, is constantly evolving. Similar to other fields, radiology practice is also currently undergoing fairly significant changes, likely due to a combination of the COVID-19 pandemic and a growing nationwide radiologist shortage.
Radiology has been defined by rapid technological advancements since its infancy with the discovery of x-rays. Radiology expanded from interpreting images of x-rays to fluoroscopy (video x-ray), nuclear medicine, CT, ultrasound, and MRI. The viewing of medical imaging exams shifted from viewing developed film over a light-emitting view-box (typically at a hospital or imaging clinic) to being viewed on computers.
As computers and the internet both became faster, sending imaging data sets to essentially any computer with a high-speed internet connection became possible. This drastically changed the practice of radiology, putting all exams (including comparison exams) at a radiologist’s finger tips. This allowed radiologists to practice more efficiently, with greater accuracy (thanks to frequently having prior imaging exams), and from essentially anywhere.
Teleradiology has experienced rapid growth since its inception over the past 20+ years. Today, teleradiology allows essentially the entire broad spectrum of all medical imaging exams to be interpreted from anywhere with a quality internet connection.
Younger generation radiologists tend to place a higher value on quality of life. This, in conjunction with a ‘buyer’s’ job market and growing burnout through the field, has led to the rapid development of “work from home” remote employee and partnership positions at private practice and academic institutions, previously largely reserved to employee or contractor jobs with teleradiology groups.
Large private and academic radiology practices have responded by creating attractive remote positions to hire within a difficult job market and help keep up with ever-increasing volumes. This trend is likely to continue and has the added benefit of taking advantage of time zone differences to improve/expand radiology coverage, particularly evening and overnight coverage.
One of the most exciting developments in radiology is the use of artificial intelligence (AI). Current AI technology can be used to help structure and organize radiology worklists, assist with radiology reporting, ensuring patients get appropriate follow-up for incidental findings, aid in the viewing of studies, and more.
Computed-assisted detection (CAD) does already exist for mammography and lung cancer screening CT, but this technology still has a long way to go to truly add value. Regardless, this is a huge step forward for medicine, and it is only going to become more commonplace moving forward.
AI is poised to drastically change healthcare over many different fronts. AI has the ability to sift through and analyze vast amounts of data and employ machine learning algorithms to find trends and subtle relationships that would otherwise be undetectable. This data can then be used to help predict future outcomes, recommend best practices, and tailor treatments to the individual patient.
Breast imaging, for example, is primed to see significant improvements in accuracy patient outcomes. Breast imaging uses a specific lexicon and grading system to evaluate breasts for suspicious breast lesions and identify breast cancer. They also have a database of biopsy pathology results for all biopsied lesions.
Artificial intelligence algorithms should be able to analyze all of this data and correlate each and every combination of imaging features and findings with the pathology.
More broadly, AI will likely be able to use machine learning to identify patterns that we humans may overlook or that are too subtle for us to recognize. AI tools will also likely be able to learn how to detect these findings on mammograms, ultrasounds, and MRI scans themselves and then classify each lesion they detect.
AI-assisted workflow efficiency tools are being developed that will help automate the tasks of a radiologist such as sorting and prioritizing exams based on turnaround time (and eventually acuity), identifying relevant images for comparison, and suggesting diagnoses through report generating software such as RadAI.
Early stages of radiology AI have already proven capable of reading/interpreting images, with some studies showing that AI can match or exceed the accuracy of human interpretation. This will augment radiologist workflows and improve our ability to read the most critical exams first, preventing delays and expediting patient care. Example – AI will be able to help identify head bleeds on CT and pneumothoraces on chest x-rays and move those cases to be read immediately.
Radiology AI is still in its infancy and will surely become an essential part of radiology in the future. Innovation, implementation, repeat.
I, not infrequently, hear concern about artificial intelligence replacing or substantially reducing the need for radiologists in the future. While AI will have a drastic impact on radiologists, it will not replace us, at least for a few decades.
AI is more likely to become a radiologist’s best friend, allowing us to work harder, faster, and smarter. Radiologists will begin to rely on the new technology and fancy algorithms that will result in greater accuracy and confidence in our interpretations, which will ultimately positively impact treatment algorithms and improve patient care.
However, we’ll be reliant on academic and industry leaders and the American College of Radiology (ACR) to help ensure that AI is used safely and ethically.
It is clear that the future of radiology is bright. Technological advances will continue to help radiologists with their workloads and improve patient care. However, there are some challenges that radiologists will face in the future.
One challenge is the growing radiologist shortage. As the population ages and more people require imaging services, the number of radiologists needed to meet this demand will increase. While radiology residency positions have gradually increased from 1,011 in 2006 to 1,113 in 2020, this has declined from the peak of 1145 in 2014.
Meanwhile, the number of practicing radiologists is unlikely to match the growing demand. Slightly over half of practicing radiologists are already over the age 55 and nearing retirement.
Additionally, interest in radiology by medical students has also been on the decline with fewer and fewer medical students applying to radiology residency programs. If interest continues to decline, it’s possible that radiology residency positions could go unfilled, which would add to the already growing radiologist shortage.
All of these factors combined will increase the workload expected for each radiologist as well as lengthen the turn-around time for radiology exam interpretation. This will negatively impact patient care as detection of clinically significant incidental findings such as pulmonary emboli will be delayed.
Another challenge that radiologists face is burnout. With increasing demands of the job (such as growing workloads/backlogs), it’s not surprising that approximately half of all radiologists report experiencing some level of burnout (percent varies by study with most falling in the 40-60% range). Some causes of burnout include long hours, night shifts, on-call shifts, growing worklist size, and working weekends.
Radiologists who experience burnout are more likely to make errors, have lower job satisfaction, and be less productive. They are also more likely to leave the field of radiology entirely, which would further exacerbate the radiologist shortage.
The future of radiology is bright – we’re talking CSF signal on T2-weighted imaging bright!
While radiologists will face challenges in the coming years, such as the growing shortage of radiologists and the high rates of burnout, we do have hope.
Artificial intelligence is poised to be radiology’s saving grace. AI will help radiologists become more efficient, which will be necessary to handle the growing workloads and help identify and prioritize exams with clinically significant incidental findings.
Radiology is vital to patient care and has a long future ahead. Though there may be bumps in the road, radiologists will continue to provide essential services to patients and help save and improve lives.