On or around April 1, 1980, someone decided the objective of an ultrasound exam was a set of images. It seems plausible on the surface. It fits the job description of ultrasound technologists, and it mirrors the role of manufacturers. From the standpoint of medical diagnosis, though, that notion departs from medical teachings going back to Hippocrates, probably even earlier to the Ebers Papyrus. I am sure the originator could not have been a radiology. I hope that behavioral psychology might provide us with an insight.
In the Beginning
The early B-Mode ultrasound settlement included physicians with a wide array of specialty backgrounds. After some initial, and very significant, technical advances, a lot of the daily clinical work across the country was handled by radiologists. Along the way, cardiologists moved up to high-speed imaging from TM tracings and obstetricians began to offer ultrasound to their own patients. Now, we have separate ultrasonic fiefdoms in multiple areas – the ER, the ICU, MSK, family, medicine, and, at the recent AIUM, there were multiple presentations and courses on dermatologic applications.
The early days were a simple time. There wasn’t much known, you didn’t have to know much, and clinical expectations were low. Acquiring satisfactory images took a long time, and they had a low yield of actionable information. The only impetus for clinical usage seemed to be avoiding ionizing radiation exposure. Most of the uses revolved around distinguishing “cysts” from “solids” for finding fluid collections of one type or another, or for looking at movements of structures in a fluid field.
I think of the decade from 1975 as the golden (and radiologic) Age of Ultrasound. There was a vast amount of academic research of imaging fundamentals leading to some major improvements in instrumentation, and there was a gigantic deepening of clinical sophistication with a rich peer-reviewed literature. There were some massive technical downsides. Image noise content was so high and data acquisition so varied from place-to-place, from operator-to-operator, and between multiple types of instruments that there was never any effective way to establish practice standards, make education uniform, or extract quantitative descriptors of tissue properties.
Nevertheless, the clinical results of the early experience with mechanical and electronic scanners were very good, and the field flourished over a wide span of diagnostic applications. An essential factor may have been due to the way radiologists handle visual data. The starting points for all imaging modalities are knowing where to look and how to look. The real work is in fitting information from images with everything you know clinically about the patient and everything you should know about what can go wrong with that patient’s population cohort. The “list” of possibilities is prioritized by the potential lethality or severity of probable conditions. The radiologist has to have an understanding of the utility of every other diagnostic procedure in his or her own facility in order to select the safest and most informative way of resolving a clinical question as a procedural sequence.
Another covert facet of image interpretation is the ability to extrapolate the consequences of a diagnosis. This might seem obvious in a fetal or pediatric ultrasound study, but it is always a factor. Perhaps this is why radiologists have been so obsessed with the pick-up rates, the sensitivity, and the specificity of screening exams ever since the days of mobile vans for tuberculosis detection. If you miss a tiny, eminently miss-able, lesion, the patient might lose years of life. Over-diagnosis has its own set of painful and costly detriments.
Fast and/or Slow
My limited exposure to “dual process theory” is from Thinking, Fast and Slow by Daniel Kahneman (Farrar, Strauss, Giroux, 2011 ISBN: 978-0374275631). This book has been referred to as a masterpiece. It has won every possible award. It explored the life work of the author and his late collaborator Amos Tversky, and it centers on “prospect theory” – the basis for Professor Kahneman’s Nobel Prize in Economics. The subject is treated scientifically. It describes many ingenious experiments, which established there are two main ways people respond to situations, referred to as “fast” and “slow.” The works of other researchers is detailed selflessly, and there are no wild speculations. Since this work concerns basic human behavior, it would seem logical that it applies to ultrasound, too.
Fast and slow are metaphors. The fast system is instant, automatic, unconscious, capricious, effortless, and always on. It is triggered by unconscious perceptions and works by ingrained associations and patterns that are very hard to change. It is incapable of calculation. It is not influenced by statistics or objective reality. It is gullible and can be misled. Effective ads appeal to the fast system.
Slow is rational, conscious, suspicious, and very effortful, because it involves a lot of energy expenditure. Fast can be brilliant, but prone to systemic errors; slow can be thorough and plodding, but it is not “perfect” either. Slow is mostly off, and it can be derailed by emotionally-tainted fast perceptions, as well as a limited knowledge base.
I’ve always believed in love at first sight. That is an ultimate fast system response. Fast is very efficient, and it works by a system of heuristics. “Heuristics” is a relatively new word, coined from a Greek root related to discovery, so its definition remains somewhat pliable. I encountered the term in college (it’s probably common in grade schools now) in issues related to computer searching, pattern recognition, and artificial intelligence. An heuristic is a fast, efficient, down-and-dirty shortcut for getting a workable and/or reasonable, approximate solution to a complex, sometimes analytically insoluble, problem. In Thinking, Professor Kahneman identifies several classes of heuristic that the fast system relies upon. One is the “halo effect” in which your impression becomes generalized over the object, i.e. love at first sight = everything about the object of your affections is lovable and perfect. Heuristics are mental habits. It is also via heuristics that biases emerge as influencers.
A Detour to Psychiatry and Genetics
Fast has been proposed as the evolutionary default state. Each of us has a balance point between fast and slow in our lives and work that I want to explore a little more. People who are locked into either of these operational states exclusively have well defined forms of psychopathology. People with very different balances between fast and slow cannot communicate very well. People who are mainly fast double down on their opinions, even when they have no factual substance or foundation, and mainly slow people cannot understand the emotional fervor of preferential fasts. All of my patients are referred, and to tell you the truth, I have always found a lot of remote referral patterns to be somewhere between rigid and irrational. I would guess these referrals are fast responses by practitioners who don’t know a lot about ultrasound, don’t keep up with technical advances in the field, and often resist informed suggestions about effective utilization of resources.
Let’s start with the common expression “Crazy runs in families.” This is true, but it has been very difficult to clarify because a family tree peppered with psychoses has so much variability by type, severity, and age of onset that their occurrence can seem random or at least unrelated. Genome sequencing has identified multiple loci for a spectrum of psychoses in which the specific whose combination of genes, and their penetrance seem to explain those variations. At one end of the spectrum is potential brilliance, the other hallucinatory divorce from reality. Think of the phrase: “She’s as pigheaded as her father.” Doesn’t it seem likely that the balance between fast and slow is also coded into our genomes?
Postgraduate Medical Education
How has medicine coped traditionally with these unrecognized fast/slow issues? Take a bunch of young people with good hearts and stellar academic records. They have altruistic heuristics, learn well, and adapt to variability and chaos. This is an ideal, new medical school class. They receive a progressive increase in information over several years to nourish their slow systems. But, there is even more emphasis on interactive topics, like taking a history. This can be viewed as a way to mold the fast system for dealing, bonding, and gaining the trust of new patients despite first impressions on both sides. It also creates an indelible bond with our professional ancestors who have all had to cope with the same issue.
Medical specialty training expands upon integrating the two systems in some way. I look back in awe, admiration, and fondness to my time as a diagnostic radiology house officer at the Massachusetts General Hospital (MGH). I was probably not so sanguine at the time. I presume the educational goals of all diagnostic radiology training programs are identical everywhere. I have never had any need to inquire of colleagues about their own background, because of the similarities of our perceptions and work habits.
Reading plane films started out as a slow system endeavor. You try to look at every detail of all of the views you have before you. It’s very tiring, exhausting, actually, and that’s before you even start to integrate clinical information and narrow down diagnostic possibilities. You keep hearing phrases like, “get the Gestalt” without knowing what that means until, magically, you get it. There are withering comments for errors, in public at conferences and more privately when reviewing a board full of cases. There is scarcely anything positive for a good call. It is much milder than the surgical experience. The system is geared toward emergent decision making and directed towards avoiding errors, and if an error is made, to be sure that it is not repeated.
The only way to handle a large volume of imaging studies efficiently is by identifying any anomalies in any part of a film at a glance. The slow system does not get evoked unless the fast system signals it needs to be activated. To do that effectively, the fast system has to be able to cope with all kind of films, with technical factors, including artefacts, and with a full range of normal variations. The immediate correlates are anatomical; the inferences are pathophysiological.
Plainly, radiologists are marvelous, especially at radiology. Every specialist has gone through a similar kind of education in their own fields, but because so much of fast system training is not conscious, you cannot relate to alternate ways of instant processing in other disciplines, even if they all share scientific foundations. You may know the end result of someone’s clinical work, but not the way he or she got there.
There has been a progressive ultrasound procedure drain from radiology into other fields that have not had the years it takes for fast system retraining for medical visual information work. It can succeed, but usually for specific questions with simple yes and no kinds of answers. It obviates the general diagnostic utility inherent in the method and the nature of its form of tissue mapping. The fast system response of an unfortunate number of radiology departments to turf issues seems to have been to relegate ultrasound to the cabinet of curiosities and to move on in other imaging directions.
Interpreting Outside the Box
Articles about ultrasound without images are like desserts without carbs. If you’re like me you’ve already looked at the image. See anything interesting?
Figure 1 transverse view of the right kidney in a 34 week equivalent infant. There is low pressure fluid distention of a major calyx and discrete reflective patches in a papilla (image courtesy of Dr. Jason Birnholz).
I selected an image from a recent visit to a level III NICU. Among ultrasound’s advantages is that high resolution, high contrast imaging can be performed in the isolette without disturbing the endlessly fragile and vulnerable small premie. Studies tend to be on demand when there is a suspicion of a problem. There are often no baseline views for comparison or pre-emptive screening for early diagnosis.
The renal image was from a female premie identified as a “normal” control, without any other information provided. For the time being, go with this one view and assume there were similar appearances for a few other papillae on both sides, but nothing else. The fast system says: yellow alert, something unusual and unexpected, presumptively pathological. There is also a vibe that the problem is local and that its cause may have been a drug side effect. The slow system cannot go much further without a lot more information, starting with why this child was delivered early and whether there may have been hydramnios. Then, it will want ALL of the available clinical information, and it may want to review what is known about renal pathology in newborns. Are the image findings predictive of nephrocalcinosis or predispose to papillary necrosis? Or will the appearance revert to normal with the accelerated healing of fetuses and newborns? There is not any available data to know the significance of the finding for renal development in early childhood or function in adolescence and adulthood. There is a nice review of high resolution ultrasound of the pyramids which raises the same concerns by A Daneman et al, Renal Pyramids: Focused Sonosgraphy of Normal and Pathyologic Processes in Radiographics.
One of the main research areas of Thinking has to do with how people make investments. What struck me was the framing of the concept with a strong condition of “loss aversion.” The first aphorism of Hippocrates states something like: Whatever you do, don’t make things worse. Radiology has the operative dicta: Do not miss anything in an image. Do not fail to provide information to contribute to a therapeutic action plan. We are all really risk and loss aversive.
Kahneman and Tversky found people faced with the same test problems may act to gamble one time and not to another, depending on the way the problem is phrased, as well as their moods and biases at the time. In addition, investors decided to “go for it” or “back off,” depending on assessments of luck and expectation of rewards. This is the fast system at work – biased, not quantitative. In our routine work, the equivalent might be the way an exam is conducted for a happy situation like a normal pregnancy with a goal of wellness confirmation versus the bleak scenario of staging an invasive carcinoma or searching for metastases. Very busy work schedules, limited patient contact time, and overly focused exam goals, promote reliance on habitual fast thinking.
I think of the alleged April Fools’ Day notion as witty product of someone’s fast system. It has spread like a plague through a vulnerable population, insuring commercial and administrative survival of ultrasound, but freezing progress or at least dampening the urge to continually improve what we do. Consider this - ultrasound and genetics are scientific contemporaries. On one side, we have unlocked our genetic code. On the other, we are mired in qualitative fluid-solid distinctions for a lot of what we do. It could be that is all we can contribute, but I hope not. It’s time we press our reset buttons and learn to invoke on our collective slow systems for all of our clinical activities.