Integrating point-of-care manufacturing into radiology

Dr. Anish Ghodadra

Dr. Anish Ghodadra is assistant professor of Interventional Radiology at the University of Pittsburgh and Medical Director of the 3D Printing Program at the University of Pittsburgh Medical Center, which established one of the nation’s first in-house 3D print labs in a medical center, producing more than 200 surgical models a year. With a background in engineering Dr. Ghodadra has a focused interest in using medical innovation including medical device design, 3D printing and image analysis to improve clinical care. He was a speaker at RSNA 2016 on the topic of 3D printing quality control and has been a contributor to ABC News Medical Unit. PARCA eNews spoke to Dr. Ghodadra about the integration of 3D printing into radiology, how it impacts the radiologist’s workflow and the role of PACS administrators in 3D printing.

Q. Do you look at 3D printing as a next step in the evolution of radiology or as something that is being thrust upon the radiology department?

A. I see 3D printing as something that could become a new modality for radiologists. Currently we've had a few revolutions in radiology when it comes to imaging of patients. First it was x-rays and then there were CTs and MRI, and then we were able to do 3D visualizations. And now we see 3D printing as being a new tool for the radiologist. Ultimately the job of the radiologist is to provide diagnostic anatomic information to referring clinicians, and while 3D printing may not necessarily be useful for the diagnosis of disease, it is a valuable tool for communicating important information about a patient's disease to a surgeon in a way that you just can't with traditional methods.

Q. A recent survey I saw showed that about 100 hospitals have set up centralized 3D printing departments in what are referred to as point-of-care manufacturing. Do you see that as the wave of the future or as something that will be limited to large institutions?

A. Actually, I do see point-of-care manufacturing as being essential for not just large hospitals but even for medium-sized hospitals. While 3D printing isn't necessarily a slam-dunk application for every possible surgery out there, there are several bread-and-butter operations to where models would be helpful. So I think initially this will be like most innovations in medicine. The academic centers are kind of the incubators of new ideas that slowly spread to non-academic places. There are already non-academic radiology departments that have their own 3D programs.

Q. Do you think the advantage of having 3D printing in-house is that the process requires the close collaboration between the radiologists, surgeons and the technicians who have to put the whole thing together?

A. Absolutely because 3D printing in radiology for medicine is essentially a small manufacturing operation where you're essentially making a custom widget for a referring clinician and this device that you're creating is a lot easier to make when you're able to just walk down the hall and talk to the radiologist or the engineer the CT technologist who is working in the lab and you can tell them as a clinician or as a referring physician, “I'm worried about disease X, Y or Z,” or “I'm worried about you know, this piece of anatomy or that piece of anatomy and I need the model to show me this.”

Being able to have those discussions face-to-face where you can all sit down and look at the imaging and decide what you need to show I think is invaluable. Without having that I think it becomes a very depersonalized application.

Q. Currently most of the applications for 3D printing are in making surgical models. Would you see it someday being more commonly used for implants?

A. Yes. I think that the evolution is going to be, we're going to start with anatomic models for preoperative planning. The next step is to move to patient-specific surgical instrumentation, for example surgical cutting guides, or drill guides for various surgeries. The logical next step then would be moving towards patient-specific implants. So first I can get a sense of the anatomy with this model. Then I can use these custom cutting guides to perform the procedure and then rather than forcing the body to mold to a premade object, we can actually create a device that is custom-made for that patient.

Q. How do you see this affecting radiology's workflow?

A. That’s a good question. So I see this as being having kind of two impacts one is the obvious one that we're adding a step at the end of the radiology workflow of creating a 3D printed model on the back end. So I have a set of images and I need to add into the workflow steps to turn that into something I can print.

But also being radiologists we can also come in a the beginning of the process. We know that without quality imaging you're not going to get a quality model. And so the technologists the PACS administrators, radiologists all have to work together to ensure that the type of imaging that is done is optimized for the production of 3D printed models.

Q. How do you get from the DICOM image to the file that creates the 3D model? Who does that conversion? And how long does that take?

A. Yes, there a few key steps in the creation of 3D printing model, about eight overall steps. First you have to have high quality imaging to start with. In general you want as thin of a slice as possible, ideally one millimeter or less. After that the next step is a process called segmentation. In segmentation you are delineating the imaging voxels that belong to the anatomy of Interest. If you're printing out a model of the kidney, then you're actually using a set of semi automated tools tracing out the individual areas. For example, on these slices this is the kidney, on these slices this is where the tumor is you're marking out where the arteries are for the collecting system, if you're working on a vascular model you're marking the blood vessels.

Q. Does the radiologist do that or the technician?

A. It is a combination. For simple models, we actually have engineers and technologists that perform the segmentation for making a model the splenic artery aneurysm a technologist or engineers able to be trained on that one application. If it's a more complicated model, let's say a model of the temporal bones or the model of a complex tumor that's invading surrounding arteries and veins and nerves, then a radiologist would actually have to step in and perform a lot of that because they're the ones who have the detailed knowledge of the anatomy.

And the segmentation process is probably the most expensive time-consuming part of the process the fastest it would go as maybe 15 minutes, but can be up to three or four hours for the segmentation for complex models. Regardless of who does the segmentation, a radiologist always reviews it to confirm accuracy. So Step one is segmentation that turns the imaging voxels, the stack of 2D slices and turns them into a 3D object.

The next step would be the CAD work, the computer aided design work, where you're taking that 3D object you've created and you’re cleaning it up, you’re trimming away things you don't need, you’re hollowing structures that are hollow and you’re adding labeling to the model, and that process is usually done by either a specially-trained radiologist, or a technician or an engineer.

And then after the CAD design work comes the actual preparing of the file for printing. Just like buildings that need scaffolding as they're being built, these models need scaffolding as they’re being printed and how you print it and how you design the scaffolding can affect the success and quality of the outcome.

Then comes the printing and after you’ve printed there remains some cleanup work that happens afterwards and then quality control and then the model itself is ready.

Q. How does the PACS administrator or the PACS staff fit into that whole scenario?

A. I think right now, on it's the most basic level, there needs to be a mechanism to allow the 3D print lab to gain access to the DICOMs, for example, we’re running a DICOM server for the 3D print lab and images are sent from our PACS to a DICOM server. And then we have you know, you have to be able to maintain data integrity and redundancy of the data and make sure all the connections are there and secure so that you can safely, easily and reliably get the DICOMS to the print lab. And then in the short term making sure that the IT infrastructure of the lab itself is there that you have a dedicated computer for storing your DICOMs and for doing all the processing work. In the long term there will eventually be a DICOM standard for the 3D files themselves, where they're actually may be a DICOM encapsulation of the STL 3D printing file.

Q. So STL will be incorporated into DICOM?

A. I believe so. There is a DICOM working group 17, I think, that is working on the integration of STL's because I can see down the road where you need to have some record of what you printed. And so I can see the segmented data and 3D data and pictures of the actual model being uploaded to PACS as part of the medical record.

Q. So does the PACS administrator then have more responsibility for helping with that conversion or is it just more storage space?

A. I think PAC administration part is more about making sure infrastructure is there to allow the 3D print lab to do its work. It is somewhat different infrastructure than what is needed for just viewing images. You need to be able to get to the raw DICOMs. Normally the images that are sent to PACS are say 5 millimeter thick slices of the abdomen, but for 3D printed reconstructions, you want to send the .625s for example, having a mechanism when a patient is scanned, while we may not send the thin slices to PACS you want to be able to get to the thin slices.

Q. Are those types of files especially for 3D printing, are they large files extremely large?

A. They are smaller than the corresponding imaging files. So size is not an issue.

Q. Where do you see radiology and PACS five years from now in terms of 3D printing?

A. I imagine that in the next five years of the two are fully integrated to where the receipt of the DICOMs for the medical images the construction of the 3D reconstruction of the anatomy of interest, and then also the storing of the 3D data all being done within PACS all in a DICOM format. This also includes not just 3D printing but also augmented reality and virtual reality because the file interfaces will be the same for both.