CME Radiology Grand Rounds

High profile challenges in machine learning and artificial intelligence such as the Jeopardy challenge in 2011 and Google Deep Mind’s defeat of the world champion at the board game, Go, have resulted in unprecendented “trash talk” about the end of radiology. Prominent articles in the New England Journal and JACR by Ezekiel Emanuel in recent days have gone so far to proclaim “in a few years there may be no specialty called radiology”. Stanford’s Andrew Ng suggested in The Economist that a “radiologist may now be in great danger of being replaced by a machine than his own executive assistant”. This comes after a prominent West Coast start-up proclaimed an end to the “wasted protoplasm that represents the radiologist at the workstation”! In this presentation, Dr. Siegel will discuss where we actually are today with articial intelligence in medicine and specifically in radiology, where we are headed in the near future and whether radiologists can expect to be replaced within the next 20 years. He will also discuss how he believes machine learning will evolve in radiology and how we can harness its power today.

The combination of the American Recovery Reinvestment Act (ARRA), the Affordable Care Act (ACA) and Medicare Access and CHIP Reauthorization Act (MACRA) are the most sweeping change to how healthcare is regulated and financed in a generation.  Together, these policies are leading to change, disruption, investment, and innovation across the US healthcare system.  This presentation will discuss the policy rationale for these related programs and discuss the implications of the upcoming election on how they will evolve and impact physicians.  We will also discuss how the private sector is responding to these policies and economic incentive changes with a focus on where new entrants are gaining traction.   

VIDEO NOT AVAILABLE DUE TO CONFIDENTIAL DATA

This Etta Kalin Moskowitz Lecture will focus on the health and vitality of radiologists, from those in training to those at the most senior levels.  Data – and experience-based information will stimulate discussion of and solutions to challenges in radiology.  These challenges include career and financial uncertainties; competing personal and professional values; changes in technology; communications gaps across continent, generations and genders; and time demands and fixed schedules.  Research findings will include interventions and Stanford and elsewhere that address values, setting limits, resilience, attitudes, services, group meetings.

The per capita radiation dose to the US population has increased by a factor of 6 since the introduction of more advanced diagnostic imaging. In addition, concerns about patient safety and misadventures involving the use of radiation in diagnostic imaging studies have further focused attention on radiation safety. Some authors have calculated the number of radiation related fatalities caused by diagnostic imaging using the controversial linear no threshold hypothesis and without taking into consideration any risks associated with not doing the study. Fairly communicating the radiation risks and any risks related to not doing a study is very difficult.  Furthermore some have recommended obtaining informed consent and tracking cumulative radiation dose as a way to manage radiation risks but these approaches may have unintended consequences. Other risk mitigation strategies (lowering radiation doses by modifying technique, avoiding unnecessary studies and avoiding repeat studies) need to be emphasized. 

There is great promise for quantitative imaging in clinical oncology. There are also many challenges and potential pitfalls. The role of imaging and particularly quantitative imaging varies in large part based upon the specific function of the imaging study. Tumor identification, characterization, staging and assessment of therapeutic response are the main roles of imaging and all these can be approached quantitatively. Ultimately, the goal is for quantitative imaging to add precision and accuracy to the imaging study. However, the variability and precision of the image acquisition must be considered as well as the computational techniques used to quantify.  Quantitative imaging may be used on the patient level for decision support, but also to correlate and link with other quantitative biomarkers such as genomics and proteomics. Incorporation of imaging into “big data” analysis necessitates the use of quantitative imaging. As such, the great promise of Precision Medicine emphasizes methods that provide quantitative resulting. Therefore, we must modify the radiologist’s classic approach to image interpretation, in the form of an unstructured clinical report and provide our output quantitatively. Ultimately, we also hope to derive other new information from the imaging study. Quantitative approaches such as “radiomics”, can extract features of a segmented tumor such as shape, texture and intensity distribution to further characterize a lesion.  To accomplish the tasks and goals, the Radiology community must identify clinical and research paradigms which could benefit from quantification. Further, software tools and database tools must be developed to capture and analyze the data, be it on a single patient level, an entire clinical trial or a population cohort. The development of the technology as well as it’s precise clinical applications truly must evolve together to optimize imaging as a quantitative biomarker.

Radiology education, no less than medical education broadly speaking, tends to be excessively focused on the known.  Questions are posed to which the answers are already known, and learners are evaluated by the degree to which they produce the "correct" responses.  Unchecked, this approach to education can prove highly problematic.  In this presentation, we review the reasons for paying more attention to the unknown.

Quantitative Imaging (QI) will be the foundation of precision medicine in future clinical care. QI concepts are presented and contextualized through the lens of ultrasound elastography. Technical advances to drive QI in ultrasound via improved instrumentation and image processing are presented, and new demands that QI will place on Radiologists are discussed.

Aspirational characteristics for both leading and being members of improvement teams will be discussed.  This includes:

1. Be a Good Listener
2. Effective Communication - Accountability Cycle
3. Simplicity: Prioritization and Pace
4. Expend Energy to Optimize People Development
5. Lead with Optimism 
6. Create a Culture of Self-Improvement, Wellness, and Sustainability
7. Progressive Attitude Towards Failure
8. Humility over Arrogance 

Human performance will also be discussed as pertains to management of teams.

Conventional MRI has been and will continue to be instrumental in the diagnosis of CNS white matter diseases including multiple sclerosis (MS); however, standard MRI contrast mechanisms (T1- and T2-weighting, T2-FLAIR, etc.) ultimately fall short of the requisite sensitivity and specificity to monitor disease status and guide the development of novel therapies. In MS, for example, the disease process is not limited to the plaques visible on T2-weighted images or even to the white matter exclusively. Rather, the prototypical demyelinating plaques of MS occur on a background of more diffuse disease that includes axonal degeneration and gray matter involvement, occult to standard imaging methods. Failure to identify and quantify these processes confounds efforts to monitor disease progression and develop novel therapies aimed at neuroprotection and myelin repair. This presentation will review efforts over the past decade to develop quantitative biomarkers of white matter disease and repair using novel MRI methods based on T1-/T2-relaxometry, diffusion, and magnetization transfer phenomena.  

Specialists may leverage several strategies when seeking to manage population health.  For radiologists, reducing variation in the imaging examinations that we recommend and how we report key findings has the potential to support more uniform and appropriate care at the population level.  Under-utilization of medical imaging risks decrements in the health of our population while over-utilization leads to increased cost and heightened morbidity from unnecessary follow-on imaging and interventional procedures. Moreover, increased precision in the quantitative nature of our reports promises to yield more effective treatments as therapies are personalized to precise patient phenotypes and disease states. Appropriateness criteria and referral guidelines take the guesswork out of which tests to recommend, and imaging-based care algorithms narrow the range of recommendations that referrers may receive in response to a clinical imaging scenario. However, such changes to our practice threaten the ‘art of medicine’ where intuition plays an important role in establishing diagnoses and understanding disease severity. Art can take many forms, ranging from the cubism of Picasso to the realism of da Vinci.  The transition from personal impression to consensus and fact-based conclusion in the tests we recommend and the reports that we generate mirror the transition from abstract art to photorealism.  The increase in precision does not make ‘art’ any less artistic; rather, it is simply based on a different set of principles.  In the era of population health management, the art of medicine is not dead -- it’s just different.

X-rays, as well as other forms of electromagnetic radiation, are subject to several physical effects including attenuation, refraction and scattering. So far, attenuation is the only source of contrast in medical X-ray imaging including plain radiography, fluoroscopy or computed tomography (CT). Imaging techniques, which are capable of exploiting other physical phenomena, are subject to intensive research efforts, since they might provide additional and potentially complementary information to attenuation contrast imaging.

Phase-contrast and dark-field imaging are novel X-ray based techniques that provide enhanced soft tissue contrast. Early results indicate that X-ray phase- contrast (PCI) and dark- field imaging (DFI) may provide improved soft-tissue resolution compared to conventional absorption CT by exploiting phase-shift and low- angle scattering, respectively, rather than absorption information. The dark-field signal is highly sensitive for calcifications and especially microcalcifications, e.g. in breast imaging.

Any phase-shift occurs when photons pass through different components of material. Most clinically-motivated PCI studies are carried out with monochromatic and brilliant X-ray sources at large-scale synchrotron facilities. Recently, by applying a grating-based interferometer, the modality has been successfully transferred from synchrotron sources to conventional laboratory-based broadband X-ray tubes. Recently, prototypes for imaging large animals such as pigs were introduced which were also employed in human cadavers.

While synchrotron sources offer highly brilliant and monochromatic x-ray beams, their high spatial and financial demands make their clinical use impracticable. To close this performance gap, new types of x-ray sources have been investigated. One possibility are compact light sources (CLS) based on inverse Compton scattering, which provide a quasi-mono-energetic, tunable and partially coherent x-ray beam.

Within the framework of the excellence cluster “Munich Advanced Photonics” funded by the German Research Council biomedical applications of phase- contrast and dark- field radiography using synchrotron, compact light source and grating-based were applied using anatomical specimen as well as  ex- vivo and in vivo experimental animals. Promising results could be achieved in lung, breast, musculoskeletal, vascular and brain imaging. Early detection and more specific characterization of emphysema, lung fibrosis, breast cancer were the most promising results of these studies. The potential and limitations of these technologies concerning clinical translation will be discussed in this lecture.