Transitions From the Origin of Radiology Into a Future With Artificial Intelligence

In 1895, Wilhelm Röntgen discovered the first X-rays. Film would later be introduced in the late 1910s, and the first contrast agents used in the 1930s. While some of these early innovations expanded the capabilities of radiology, others were displaced by additional discoveries years later.

Conventional radiography on film eventually became obsolete to digital radiology on computer systems. Issues with digital radiology due to modulation transfer function deficiencies and lack of access to high-resolution displays were soon minimized. Additional features allowed radiologists to interpret image findings, utilizing features such as window/level contrast adjustments and magnification.

In the 1970s, Sir Godfrey Hounsfield, CBE, FRS, designed the principles of CT scanning from a three-part paper on the technique of computer transverse axial scanning. During the 1990s, sometimes called the golden decade of radiology, CT scanning began as a cancer screening tool and evolved into a complex measurement tool for the acute patient. A critical factor influencing the favorability and evolution of CT scanning in radiology was the transition from invasive diagnoses and treatment to minimally invasive or even noninvasive diagnoses and treatment options.

After nearly 50 years since the development of CT scanners, the future of radiology has now pivoted in many ways to an era of data and artificial intelligence (AI). Based on the extent of imaging data collected digitally, many AI algorithms cleared by the U.S. Food and Drug Administration are radiology focused. Other organizations, including the ACR®, have instituted programs, such as the ACR Recognized Center for Healthcare-AI (ARCH-AI), to further certify these programs.

AI-assisted radiology is one component of this new revolution, poised for augmenting radiology workflows and further improving the field. This may benefit radiologists and enhance image analysis, improve clinical decision making and advance image rendering and reconstruction.

Overall, the belief is that these systems will provide increased accuracy and expedite diagnoses, positively impacting patient outcomes. Currently, large language models may incur harmful errors, but eventually they could produce radiology reports in laymen’s terms for patients.

As with every field of medicine, patient populations are growing, and the degree of work has increased exponentially. Many of these AI programs are focused on lessening this burden, but others may lead to the development of new technologies that could displace current imaging techniques, lessening exposure to radiation, increasing contrast between tissues or even expediting the time required for image collection and processing.

Overall, as these new technologies take shape in the clinical environment, their impact on radiology in the future remains unclear. Although perhaps ominous, it also represents countless opportunities for aspiring radiologists to shape and guide the development and implementation of these tools to ensure superior patient outcomes and safety.

While the extent of AI adaptation into practice is cloudy, what is clear by recent technological developments is the inevitability of its utilization in radiology.