Mary Morrison Saltz, MD

An odd tumor of the pancreas, held in the hand, a death - the history of watery profuse diarrhea. Why, what, how?  That was the gross pathology of discovery, a window to disease.

The year was 1958, the pathologist, my father Duke, 

 in medicine to the world. A discovery made from the history, and the course of the disease, combined with information gleaned post mortem-holding in hand the organs, weighing the pancreas, observing the masses and finally looking under the microscope

A syndrome described. Verner-Morrison syndrome - yet what was it and why? 

, pancreatic cholera. An islet cell tumor. Vasoactive intestinal peptide was discovered, a molecular cause determined. Now the gene sequence is analyzed.  Going deeper and deeper into the root cause of disease is what we do, and when we think we might understand, we realize there is yet another layer of knowledge to gain.

Today we stand at the intersection of genetics and radiology: radiogenomics. We are the de facto gross pathologists of the age, wielding our imaging scalpels to cut through the body swiftly to diagnose, to treat and to follow.

As medical knowledge explodes, we are there to discover, to guide and to direct therapy. Often it is we who first discover a mass -but is it cancer or not? We use now tried and true guidelines to decide further action. Is that pancreatic tumor big or small, cystic or solid, enhancing or not, single? Multiple? Encasing the vessels?  

With this knowledge in hand, decisions are made and care given, often a biopsy obtained. If we don’t think the results are what we expect, we say go back, get more tissue, do it again! What if we know not only the cell type but the genetic expression of the tumor? Could we help more? Give better guidance?  

What if we could combine the big picture we see with ever more specific data, delving into not only the pathology but beyond, into specific expressions of tumor genetics? Not only to see the gross pathology but to combine magnetic resonance spectroscopy, molecular imaging, microscopic pathology and genetic information. What a powerful tool this could be - taking imaging and cancer treatment to the next level, many steps beyond Virchow, beyond WDHA and into the new world of radiogenomics, correlating genetic information with radiology images.

Imagine starting with a breast MR, swooping down to see the digital pathology slide, and yet again through higher and higher powered resolution and more - down to the sub-cellular level.

A Google map for the body. Yesterday we had gross pathology, a surgical specimen or an autopsy, yielding an organ to hold in the hand, to weigh and to dissect. Next the pathologic glass slide, carefully prepared, stained and ready to examine. These findings, in the setting of the patient’s history, their signs and symptoms, gave us the basis to diagnose and treat. 

Today we have gone digital - and have replaced the touch and feel of the organs with the dynamic non-invasive gross pathology - the CT scan, the MRI. We have added functional molecular imaging, MR spectroscopy, fusion imaging. The virtual microscope, 

, gave us a digital version of the glass slide. 

Today this technology allows us to zoom in and out, and to move through a block of tissue, virtually.  Now we are able to add the genetic information, in a “mark-up” - placing it in three-dimensional space.  We know that combining genetic information about tumors and their radiologic appearance adds 

 and that MR spectroscopy is more powerful when combined with 

This is the beginning of an information explosion, and we must prepare as imagers and physicians to lead our medical colleagues through this journey.

As those most able to visualize and integrate imaging data, radiologists are uniquely poised on the edge of the new imaging frontier, where placing genetic information in three-dimensional space is now possible, disease rediscovered, combining information across modalities, then going down to the next level of resolution, then the next, will yield a wealth of knowledge.

Using our current understanding of disease, based on centuries of science, while leveraging state-of-the-art advances, will bring us to the cutting edge of patient care, providing novel and personalized therapies.  As imagers we will be able to see and monitor these interventions in a new way - personalized radiology of the future, here today. 

Let us seize this opportunity as imagers, and take it to the next level.  What do you think?

Articles

At the intersection of genetics and radiology - radiogenomics - we as imagers can see and monitor interventions in a new way: personalized radiology of the future. 

Radiogenomics: On the Road to Personalized Radiology

It was a Saturday night not long ago. I was sitting far from my grown baby when the electronic umbilical cord began to tug. The texts began during dinner out, interrupted, ruined by worry, and they kept coming until 2 a.m. By then my daughter had undergone her second drainage in ten days, for the same thing: a peri-tonsillar abscess.

It was weeks in the brewing, seen by a parade of ENT docs, providers at student health, and entailed three trips for me. At a distance, I was helpless in the face of the medical machine, even when it was my daughter. My first trip, I went home reassured after her endoscopy was negative. Time passed, her course worsened, and the imaging began. When in doubt, get an X-ray.

Her first CT was negative, and her doctor told her she did not have an abscess, but two days later the abscess that never was was being drained in the hospital. That was my second trip. Negative CT - really? Really?

Soon enough again she could not swallow, not even saliva, and went alone on a Saturday night to Emergency, afraid and quite ill. No more Dr. Nice Mom. My email to her attending laid out a reasonable medical approach, which I suggested he follow - or deal with me.

First, I suggested, if you want to see what is going on in her neck, a CT with contrast is imperative.  Repeating a CT without contrast, his previous test of choice, would only compound the mistake of ever having ordered it in the first place.

Of course, there was a large peri-tonsillar abscess present, easily seen, and well described in the report.  What was 

 said in the report was that it looked the same as the other CT scan, but the abscess was just easier to see with contrast, and had been missed by the first radiologist. 

Why is this our collective fault? When will we lend support to better decision making? 

, the test of choice in a patient with fever and a neck mass is a CT scan with contrast. 

No surprise, right? We all know that, but not everyone else knows that.  Not even really smart sub-specialists with great training and good hearts.

How many times a day do you come across a patient in whom the wrong study is ordered? Sometimes it is brought to our attention before the test is done, and then we take a deep breath and try to fix it.  Insulting our referring docs by telling them they ordered the wrong test is never fun. That is if you can even reach them. Where is the incentive to be scolded for bothering them? And of course then the insurance pre-certs may no longer apply. And the hospital loses money, and we don’t get paid… And once the wrong test is done, it is too late.

It is an imperative that we leverage technology to help our profession ensure the right test is done the first time.  Decision support is key. 

The ACR has been working on this for some time, by establishing criteria to rank suggested imaging exams for given diagnoses. The current tables are crude, and their impact across patient populations has not been tried and tested. 

As the EMR becomes universal and we implement computer aided decision support at the point of physician order entry, we will begin to change ordering patterns established for decades. This will not be without difficulties and many doctors will figure out how to order what they want, by working around the system. 

Yet we will gain the power to circle back and see, for each doctor and for each population studied, what makes an effective and sensible ordering practice and learn from this. Big data and data analytics will be brought to bear on what test to order when, analyzing what works and what doesn’t. This will allow refined suggestions for imaging, based on an individual doctor’s track record, based on the population considered, and based on things like seasonal prevalence of illness. What is suggested in flu season may not be appropriate in summertime.

Data analytics in concert with computer aided decision support will help ensure that the right test is ordered the first time. It will allow us to take a second look at studies where the imaging findings are unexpected, given the clinical situation.

So next time a patient comes to a specialist’s office with a suspected peri-tonsillar abscess, a CT with contrast will automatically be ordered, and if it is “negative” despite a high clinical suspicion, we look one more time to make sure we did not miss the diagnosis.

They ordered a CT without contrast and missed my daughter’s peri-tonsillar abscess. We must leverage decision support to ensure the right test is done the first time. 

The Night Decision Support Came to Visit

Meet the data entry clerk of my medical school era: A large and ponderous person slowly entering the written orders into the computer terminal. The clerks sat guarding their terminals, protecting unauthorized access, scaring away all but the most intrepid.

These women, in their implacable solidity, were unlikely representatives of the future, but were indeed the bridge from the handwritten to the world of the EMR and CPOE. Without their support, your orders never got entered, and as such each one was my best friend on every nursing station.

This was the beginning of meaningful use, the first baby steps to today.

By its very nature, medicine is an art, complex and fully nuanced. A science, yes, but more than that. The interface of medicine and the computer is a test of the information age. To reduce our art to a binary system of "yes" or "no" remains a challenge despite the great advances of the information age.

Consider something as simple as a belly X-ray: KUB, flat and upright abdomen, abdominal series, flat plat of the abdomen, abdominal film, r/O free air. The same? Different? How to enter? This is still a challenge for CPOE.

When a person read the doctor's order and entered it into the system, they knew a lot about what the doctor actually wanted and what they really meant by their order - and usually they were right. They translated the written order and served as a human interface to technology.  Since then we have made progress, but even how to integrate all the data we have today is uncertain. And more data is coming soon.

Discrete data elements, data mining, big data, clouds, meaningful use - what can that mean to us as radiologists? In 2014, as 

 rolls out, we will have a tsunami of data. How will we use all that data to improve care? Can we take baby steps, or jump in feet first to the storm?

Americans are big spenders on healthcare with poor results, as compared to our peers around the world, but how to do it better?

Can we as radiologists make sure to be involved, and create a world where more information is better and helps us do our job better? Something as seemingly simple as making sure the right test done right the first time? In the ACO model costs will be cut. We must meet this challenge, and with fewer resources deliver better care. Can we?

What does this mean for us? As radiologists we have to ensure that the right test is done the first time, make the correct interpretation, and deliver the reports to the right docs and ultimately to the patient. Knowing more about our patients will be imperative. More and more data, shared and available from millions of patients, will allow us to deliver better care. 

 and using it to improve care delivery is a pressing challenge today facing radiology.

Data analytics will allow delivery of personalized radiology.

With a personalized approach, we will be able to recommend the right tests, and have the data to back us up. This might mean suggesting that one patient not have a breast biopsy, but rather an MRI. Then at once we could communicate this suggestion, not only to the referring physician, but also to the patient herself - directly and quickly, via patient portals.

For another patient, perhaps superficially similar, we might suggest routine follow-up, and a third, with the same imaging findings, might get an immediate tissue sampling. Looking at large scale data using sophisticated data analytics allows a more complete and confident conclusion than does looking at one individual in isolation. We will be able to not only integrate this patient's data, but that from similar patients, looking from things as simple as their zip code to complex genetic information, and make a recommendation specifically tailor-made for that individual - indeed 

Using the power of data analytics, the cloud and big data, we will be able to better direct and monitor patient care.

By tailoring imaging to meet the needs of each patient and strategically selecting imaging care plans tailored to each individual, I believe we will reduce costs, reduce unnecessary radiation, protect ourselves from ligation, and allow patients to be increasingly involved in their own care, with more patient-centered radiology.

The growing amount of healthcare data and the use of analytics can improve patient care and reduce costs. But can radiologists take advantage?

Using Big Data to Improve Patient Care in Radiology

It was ages ago that I was standing, a little bit excited but not about to admit it, on the ward for the first time: a real medical student. My white coat was crisp, fresh, new and a dead giveaway to my newbie status. My Littmann stethoscope worn in the self-conscious style in the way doctors did, draped like a scarf carelessly over the shoulders. In my pocket: the 

. It was an ancillary brain - the source of all one needed to know in the moment. The resources of the library not available until later, if one could sneak away to seek greater knowledge, to create a handout for rounds in the morning - something to refute the dogmatic wisdom of the JAR, but never to challenge the attending. 

Such was life in the 1980s, a world just after of the era of The House of God.1

, the house staff and the attending. Google scholar at a touch. Drug doses all there, differential diagnoses at a flash. No longer could I convince the team that a vegetable fiber in a stool specimen was a worm. I would have to show them a picture of a worm, and the most loquacious of chat could not convince them without a picture. Bravado does not get you where it once did.

Today we are living in age of fusion of the human brain and the machine - in essence, the cyborg. The man meets machine era is here - now and more and more in the future.

As radiologists we live by technology and die by the same sword. Where would we be without CT or MR, never mind PET/CT, a dynamic intersection of form and function? Yet, as a profession, we are reluctant to embrace change. But change will come whether or not we embraced it.

As doctors we hold peoples’ lives in our hands. As imagers we excel at pulling a whole story, an entire patient, from a series of pictures, often black and white, sometimes in color. We recreate a real person from the images in front of us, telling a story of the athlete who fell and damaged his knee - will he play again …this season…ever?  The young mother with breast cancer, with positive nodes but no distant mets… or did we miss one?

Statistics are scary. Some studies say that up to 30 percent of our reports contain errors2.  

 sites a much lower number - around 4 percent. Yet if you think how many studies we do every day, either percentage is a lot of mistakes - mistakes that are often obvious to us in retrospect, so obvious we can’t believe we missed them. One former colleague  who was sued for missing a one-centimeter tumor on a CXR, gave the defense that he never read the case. The finding was so obvious he could not have missed it.  How do you think the verdict went?

We all agree that we want to deliver the best care to our patients, submitting a reliable report - not usually, but without fail. As we strive to approach this goal, we need help, because we are not even close to perfect. Yet in radiology, a field defined by technology and one that would not exist without sophisticated scanners and amazing imaging technology, we are slow to adopt computer aids in our day-to-day life.

What about CAD? If we wait, saying that CAD stinks and we don’t like it, it will be developed in spite of us, but not in an optimal way to help us deliver the best possible care. To ignore the changing world around us would be like eschewing Google for the Dewey Decimal card file. The challenge is even greater for those of us in the private practice world, where residents are not there to smooth the bumps of imperfect technology, allowing early adoption of even suboptimal advances.

Where is CAD today?  Well, for sure it is not here yet. Despite widespread adoption of CAD in mammography it has not been embraced enthusiastically, and although it may increase cancer detection, this comes with 

. In the lung - for sure not yet. Colonography? Pretty pictures for sure, and help some, but a 

Men and machines process information differently, and together they may provide more than the sum of the parts. Take that momentous day in 1996 when a computer trumped a chess master. In answer to that failure of man, the chess grand master Kasparov created advanced chess, a game that allowed players to use off-the-shelf computer programs to play against grand masters and machines. Ordinary chess players, using this readily available technology, trump either chess masters alone or machines alone - a type of cyborg victory. 

CAD is not ready for prime time yet, but drawing on the thoughts of 

, as they looked at CAD in pathology, the future may well be around the corner. Harnessing the power of the machine-man interaction may be the key to effective implementation of CAD, effectively increasing accuracy in our radiology reports.

We are not there yet, but the days of the Washington Manual in one’s pocket are long gone, and as a profession we need to embrace technology to improve our workflow, optimize our accuracy and leave us some time in the day to enjoy our work.  Let us actively partner with technology to make sure we are driving the process, and not let it drive us.

2 Kruponski, EA, The Importance of Perception Research in Medical Imaging, Radiation Medicine: Vol. 18 No. 6, 329–334 p.p., 2000

The man meets machine era is here, and harnessing the power of the machine-man interaction will be key to effective implementation of CAD. So get ready.

Mary Morrison Saltz, MD

Articles

Radiogenomics: On the Road to Personalized Radiology

The Night Decision Support Came to Visit

Using Big Data to Improve Patient Care in Radiology

CAD in Radiology: the Cyborg Advantage