Winter 2017

A virtual model of the heart developed by biomedical engineer Natalia Trayanova and her team promises to personalize cardiac treatment and improve care. Plus: Oncospace eliminates the learning curve to better tailor radiation therapy, the power of PET, charting the future and more.

The Virtual Heart

3-D model personalizes cardiac diagnoses to improve treatment.

Irregular heartbeats, or arrhythmias, are quite common. While benign more often than not, they can also lead to deadly cardiac arrest in some cases.

But treatment for arrhythmia is invasive and not without risks. It involves threading a catheter through a vein into the patient’s chest and implanting a defibrillator in the pectoral cavity, where the device can detect the onset of arrhythmia and jolt the heart back toward a more normal rhythm.

The question of whether or not to treat is currently being answered according to a measure known as the “ejection fraction”—roughly put, the amount of blood that circulates out of the heart’s ventricles with each beat. If your number is too low, you get a defibrillator.

“The problem is that this criterion is very insensitive,” says Natalia Trayanova, the Murray B. Sachs Professor of Biomedical Engineering and Johns Hopkins inHealth-funded investigator. The devices now being implanted end up actually firing in only about 5 percent of cases per year, so “we need a better way to stratify these patients,” she says.

This is where Trayanova’s virtual heart comes in. Similar to the way Google Earth software allows you to zoom in on most any spot in the world, her “virtual heart” gives that sort of a glimpse into the inner workings of a specific individual’s heart.

The geometric model of the heart she and her team work with has a resolution so high that it reaches very nearly to the cellular level and, in fact, incorporates information about electrical processes at both the cellular and subcellular levels. This work involves quite a bit of heavy lifting with computers. From day one, Trayanova’s lab has ranked among the top users at the Maryland Advanced Research Computing Center, housed on the Johns Hopkins Bayview Medical Center campus (see “Making Its MARCC”).

Last spring, Trayanova and cardiologist Katherine Wu published a proof-of-concept study in Nature Communications showing that their computational simulations in an individualized virtual heart could outperform the ejection fraction as a predictor of trouble down the line.

Their Virtual Heart Arrhythmia Risk Predictor—used in retrospective fashion on 41 real-world cases—predicted with five times greater accuracy than the ejection fraction which patients face the highest risk of a life-threatening arrhythmia.

“We were able to create a personalized, highly detailed 3-D heart based on the patient’s specific anatomy,” says Wu, an inHealth pilot investigator. “We could do all this without requiring the patient to undergo an invasive procedure.”

If the study’s results hold up in clinical testing, many thousands of patients in the years ahead will be able to safely avoid the risks associated with defibrillator implantation.

Making Its MARCC

The Maryland Advanced Research Computing Center (MARCC), which opened in 2015, is proving to be a big hit with faculty researchers like Natalia Trayanova and Katherine Wu, who are pushing precision medicine forward.

The supercomputing facility on the Johns Hopkins Bayview campus—a joint project with the University of Maryland and financed with state dollars—is made up of 19,000 individual processors and is capable of handling 17 petabytes of data.

Such power is crucial considering, for example, that a single MRI or CT scan creates one to two terabytes of data.

MARCC’s director, Jaime Combariza, reports that the $30 million facility has been running at more than 90 percent capacity basically since the day it opened. He is already working on plans to upgrade the facility so that it retains its utility in the fast-moving world of information science, where access to such a center can serve as quite a powerful tool in recruiting top graduate students, postdocs and faculty members.

Says Combariza, “I don’t think anyone expected us to have this level of success and this much usage this soon, but it goes to show that this type of facility is not a tool for the future—it’s something our faculty needs right now, today.”

Eliminating the Learning Curve

Looking to past patients to better tailor radiation therapy.

By the time new patients come to see radiation oncologist Harry Quon, they’ve been through an emotional ordeal: the dread and fear that arises with a diagnosis of head and neck cancer, and the long run of resulting medical appointments—meeting new specialists, getting biopsies and blood tests done, and sitting through a series of high-tech imaging sessions.

In Quon’s office, they prepare for a new stage in their journey: radiation treatment, which frequently means daily visits to a clinic for six weeks. It’s Quon’s job to make sure that the marathon is worth it. He is responsible for developing a treatment plan that will deliver enough radiation to target the patient’s tumor and yet not so much as to do irreparable harm. Radiation oncologists call this target the “achievable dose.”

To help him meet that goal, Quon has come to rely on a new tool developed at Johns Hopkins called Oncospace—a database populated with 3-D imagery and other data gleaned from thousands of prior cases across four institutions. Oncospace, part of Johns Hopkins inHealth, allows Quon to tailor the treatment plan he has in mind to the details of each new patient in a powerful new way.

The focus of the process at this point is the patient’s anatomy. When Quon submits a plan to Oncospace, he is looking for cases with tumors of similar size, shape and location, as well as those that are situated in similar ways around key organs, to see how his treatment plan compares. In effect, he can learn from the experience of past patients to develop the most effective plan.

“We’ve shown and published that using this tool helps us to lower the doses that go to critical structures in the body while maintaining the target coverage we want on the tumor,” says Todd McNutt, director of clinical informatics in the Department of Radiation Oncology and Molecular Radiation Sciences. “This isn’t pie-in-the-sky stuff—we’re using it in the clinic, and it’s helping us improve the treatments we give to patients.”

First tested successfully with head and neck cancers, Oncospace is now a go-to tool in radiation oncology at Johns Hopkins for many cases of pancreatic, prostate and lung cancer as well.

The next challenge involves taking Oncospace to a level where it becomes a tool that’s useful to clinicians throughout the course of care. By adding more personalized measures of patients and their cancers, including genetic data, McNutt and Quon hope that Oncospace will be able to predict in advance which patients are going to struggle with various side effects—weight loss, for example, or losing their voice, or their senses of smell and taste.

Such predictions could take personalized care in radiation oncology to another level. If a radio DJ needs her voice at work every day, can her therapy be targeted in ways likely to protect her livelihood? Similarly, can care be tailored in ways that will allow a chef to retain the sense of taste that is critical to his work?

The long-term goal is to make individualized decisions not just about eradicating the cancer, but about which treatment tools to use in specific cases. That could drive personalized care to another level.

There is a touch of linguistic irony involved in this next stage, considering that the first step in the work involves de-personalizing patients by turning their experiences into a set of numbers that might be useful in a database. In radiation oncology and across most other fields of medicine, that level of record keeping happens now only in clinical trials, where meticulous data about the presence, length and severity of treatment side effects are necessary to demonstrate safety and effectiveness.

“That’s the real question going forward,” McNutt says. If making that level of data gathering a routine affair can help improve patient outcomes, “Why don’t we treat all patients as if they’re on clinical trials? And how do we get that done?”

One step in that direction is by facilitating direct patient input. During visits to the radiation clinic at Johns Hopkins, patients are using tablet computers to provide information about side effects and pain levels. A mobile phone application is in the works that would allow such tracking on a 24/7 basis.

“During visits, you only get a measure of what’s going on at that moment, but we are looking to get a much more continuous evaluation of their well-being,” says McNutt.

Other key issues remain a work in progress as well. The nuts and bolts of how the Oncospace database will function as a business still need to be sorted out. Complicated regulatory hurdles are ahead as well. The four institutions in the Oncospace consortium (Johns Hopkins, the University of Washington, the University of Toronto-Sunnybrook and the University of Virginia) all keep their data in separate silos, as sharing that material across institutional lines is nearly impossible under current HIPAA privacy rules. When Quon submits a treatment plan to the database, it goes through each of those silos separately, and the results are aggregated into a final report.

“This is where we are,” McNutt says. “There is still so much to figure out, a lot of gaps that need to be filled. That’s true not just here in radiation oncology, but in all of medicine.”

Charting the Future

The traditional patient “chart” that has given way in recent years to the electronic medical record began life as a glorified accounting tool. It still serves related functions when it comes to insurance reimbursements and protection from malpractice claims. It has also evolved into a communications tool, of course, giving health care teams quick access to a case.

“Going forward with personalized medicine, there needs to be a complete mind shift” in the way physicians keep records, says radiation oncologist Harry Quon. “The documentation is no longer going to be a form of communication between professionals, primarily—it’s going to be a very detailed record of clinical actions and outcomes. That is what is going to give us new tools to measure heterogeneity among our patients.”

That future remains off on the horizon, however. Current incarnations of the electronic patient record are simply not up to this task yet.

“This is still a new concept,” says radiation oncology physicist Todd McNutt. “There’s quite a difference between collecting data for new discovery the way it has been done, through hypothesis derivation and testing in groups of patients, and discovery based on information that’s captured in the course of routine care with an eye toward interventions for new patients based on the experiences of all those prior patients.”

Moving Beyond Silos

The time is now to create an informational infrastructure that can “talk and share.”

Elizabeth Jaffee, deputy director of the Johns Hopkins Kimmel Cancer Center, is an expert in cancer-related immunotherapy. Her studies elucidating how the immune system works in cancer—and how vaccine approaches might prevent and treat it—have borne fruit and inspired follow-up work in labs around the world.

So she was a natural pick last spring to co-chair a blue-ribbon panel of top experts tasked with launching a so-called Cancer Moonshot. Working under the leadership of former Vice President Joe Biden, the panel issued a report last fall that put a premium on the need to build a brand-new kind of informational infrastructure to speed the advance of personalized medicine.

Among the 10 recommendations issued by the panel are calls to build a national cancer data ecosystem, mine past patient data to predict future patient outcomes and develop a 3-D cancer atlas.

“There are computational biologists out there who look at recommendations like these and what we’re trying to build and say, ‘This is easy to do if you have all the information,’” Jaffee says. “This is totally realistic, right now. We don’t have to develop new technologies.”

The challenges on the horizon are more logistical than technical. As an example, Jaffee singles out the various systems of electronic medical records that have popped up over the past two decades and been put into place in an ad hoc way that varies from institution to institution.

“We’re building these silos,” she says. “And if we get too far down the road where every institution has a different kind of system that does different kinds of things and those systems aren’t able to talk and share with each other, it will become a lot harder in the future to turn back and get it right.”

Regulatory hurdles lie ahead as well. Here, Jaffee points to a need for revising the Health Insurance Portability and Accountability Act (HIPAA) to facilitate the development of personalized medicine while still safeguarding patient privacy.

“The HIPAA rules we have now came out of a very different time, when HIV and AIDS were just coming in,” she says. Two decades later, “patients are as frustrated as anyone” with the need to sign off on HIPAA details every time they take another step through the health care system. “It’s at a point where it can get complicated when they’re just going to hospital A and saying, ‘I need you to send my genetic data to hospital B,’” Jaffee says.

Based on her experience, most patients are willing, even eager, to participate in a future where personalized medicine is a routine clinical affair. There is, after all, an attractive, pay-it-forward aspect to the concept, with each new patient who comes along feeding a database that will in turn boost the prospects for future patients.

“People are very willing to sign off on this type of medical practice,” Jaffee says. “What we have to figure out are the details—how do we make a consent form that works for the patient and also allows researchers access to the information they need about the case?”

As this issue went to press, Jaffee was optimistic that the work she put into the Cancer Moonshot will pay off, despite the uncertain political landscape.

“Cancer isn’t a partisan issue—this should be a no-brainer,” she says. “But even if it doesn’t move forward with the federal government, there are other ways to get things done. It’s been amazing, the way this project is bringing together all these people—from pharmaceutical companies, foundations, technology companies, academia, just every sector—to try and address this. I feel like it’s already gone beyond just the government, and if we can move things forward via people working together in the private sector, it may actually move forward a little faster.”

The Power of PET

For more than two decades, cancer researchers here have turned to radiologist Martin Pomper and his team of positron emission tomography (PET) specialists for help in finding receptors, enzymes and other drug targets—and charting the progress of novel therapies—within the body.

PET can produce finely grained images of processes in the body at the molecular and cellular level, potentially identifying diseases earlier than conventional imaging or other tests.

Pomper and his lab have developed a number of imaging agents that have been used in cancer patients. By chemically bonding radioisotopes or bioluminescent chemicals to white blood cells, the scientists use the agent as a window through which clinicians can pinpoint and track changes in tumors.

“In cancer, precision medicine is increasingly used to guide management in real time in individual patients,” says Pomper, director of nuclear medicine and molecular imaging and Johns Hopkins inHealth investigator.

Much of Pomper’s work has focused on molecules that latch onto the prostate-specific membrane antigen, or PSMA, expressed in prostate cancers that are growing and spreading quickly. “During the PET scan, we can identify lesions that are a few millimeters in size, a finding that often changes management of a patient,” he says. This can help reduce the overuse of toxic cancer-fighting drugs or tissue-damaging radiation therapy.

PET imaging agents can also be weaponized to destroy cancer cells. One of Pomper’s compounds, Lu-2, synthesized by radiochemist Sangeeta Ray, has already been turned into a smart weapon against prostate cancer. Lu-2, which latches onto PSMA on the surface of cancer cells to deliver a fatal dose of radiation, is currently under license to Advanced Accelerator Applications, a French pharmaceutical company.