The essence of medical imaging lies in understanding the relationship between patterns of energy emanating from tissues and the underlying state—healthy or diseased—of those tissues. This fundamental paradigm will not change in the future.
However, the way we study biological tissues with different forms of energy and how we draw inference from image data will change continuously at a relentless pace.
From Radiographs to Parametric Imaging
For the better part of 100 years, physics was the dominant scientific basis of radiology and X-ray attenuation was the paramount measureable parameter. Radiologists spoke of “images” and “radiographs” not “attenuation maps.” New energy sources—magnetic, radiofrequency, sonic, optical and nuclear—combined with fast, dynamic, digital methods of applying and recording them, have added dozens of parameters to the imaging toolkit.
The richness of measurable parameters has taken medical imaging beyond organ anatomy and pathology into the realms of physiology, pharmacology and cellular and molecular biology. The scale of measurement has been extended from centimeters and millimeters to encompass micrometers and nanometers. Taken together these developments are moving radiology into the age of molecular medicine and genomics.
Images as Data—Derivation of Additional Parameters
Digital images are more than pictures; they are sources of data that contain important information not qualitatively perceptible by human observers.
Hundreds of secondarily derived parameters can be extracted from image data sets by advanced computational methods, such as analysis of tumor textures, that can be empirically linked to different tumor genotypes. Computationally derived images can depict information from multiple parameters allowing us to see how they relate to each other temporally and spatially. Going forward, we will still talk about images, but the conceptual key to diagnostic inference will be gaining an understanding either directly or empirically of what each parameter represents and how that parameter is manifest in a given disease process.
Radiation Dose Reduction and Phase Contrast Imaging
Improvements in X-ray based imaging in the next decade will result in reductions of radiation doses to the point where the issue will no longer be of discussion or concern.
Current calculations projecting excess cancers and cancer deaths from CT seriously inflate the risks, because they are derived from 10-year-old data that don’t take into account new reconstruction methods and scanning systems developed in the last decade that have reduced radiation doses substantially.
Phase contrast X-ray imaging is likely to be the next new imaging method to be explored clinically.
Compared to attenuation based X-ray imaging, phase contrast has the theoretical potential to reduce doses by 10- to 100-fold or more due to the inherently high contrast it affords. Predictably, it will take time to achieve these levels of benefit but the underlying physics is favorable—phase shift versus linear attenuation of X-rays in biological tissues will usher in the submillisievert era of CT imaging.
Information and Communication Systems
With the Internet, borders have blurred between the concepts of information and communication systems, making access to data and distribution of information faster and more efficient. Mobile and wearable media will accelerate these trends. Timing of information delivery will be tailored to medical need. Biometric and/or wearable patient identification media will facilitate the “electronic round trip”—automated patient identification and no reentry of patient data or selection from pick lists required from the time of computer order entry by a referring physician until report delivery.
Direct patient access to information will democratize the medical record; all physicians, including radiologists, will need to learn how to craft reports that convey necessary information without unduly alarming patients and be mindful that many patients are not medically literate. These are unsolved challenges today.
Big Data, Data Mining and Value Creation
Radiology led the way into the era of digital medicine. Now in the era of “big data,” radiology will continue to lead in mining and mobilizing data—turning dumb data into smart knowledge to be delivered in real time—just-in-time—at the point of care.
Decision support (DS) systems for referring physicians will be built into the work process for computerized physician order entry (CPOE).
DS systems will guide radiologists in their recommendations and reduce wasteful variations in practice.
Real-time data-mining during the reporting process will be used to help avoid errors—for example, checking consistent use of right versus left and comparing terms used in the body of the report versus the impression.
Standardized nomenclature based on imaging ontologies such as BIRADS™ and RADLEX™ and structured reporting will facilitate data-mining for many applications, including the aggregation of similar cases to look for new patterns in the image data or to test new imaging biomarkers for accuracy. Radiology subgroups and the specialty more generally must work together to agree on unambiguous standardized nomenclatures to avoid confusing referring physicians and each other—is it a heart attack or a myocardial infarction, a cyst or an inclusion cyst, a tumor or a mass?
Imaging in the Era of Precision (Personalized) Medicine
The fundamental principle of precision medicine or personalized medicine is a definition of ever smaller, more precise subgroups of patients with similar characteristics who have similar prognoses and are likely to benefit from the same therapies.
The term “biomarker” is used for any finding that is linked to the presence or severity of a disease such as blood pressure, heart rate, hematocrit and other laboratory values. By analogy, what we have historically called “Roentgen Signs” may be thought of as imaging biomarkers.
Conceptually, the radiology report is an enumeration of the imaging biomarker and, as such, constitutes an “imaging phenotype” at that point in time. Imaging phenotypes are systems for scoring, categorizing and classifying disease processes and their severity. They define these “precise” subpopulations.
Establishing linkages between genotypes and imaging phenotypes (radiogenomics) will serve as the foundation for surveillance of disease manifestation—occurrence, location, extent, severity—and discovery of genetic polymorphisms.
Radiologists should begin considering their interpretations in this conceptual framework if we are to take a leadership role in the era of precision medicine as productive, vital members who speak a common language.
Challenges and Opportunities
Future developments will certainly entail vastly increased complexity in imaging technology and radiology practice, and the increased educational activities those advancements will require. Competition, both clinically and in research for “ownership” of imaging methods, will continue to increase due to the high value inherent in medical information.
On a positive note, the future will bring new capabilities that have even greater medical value. We will see radiation dose continue to drop and utilization of imaging services become more efficient, with fewer healthcare resources wasted, including the increasingly scarce commodity of time—to the benefit of patients, physicians and workers in the healthcare system
