All photos by Bryce Vickmark.
Over the last few years, there has been something of a 3D printing revolution within the world of academic hospitals, and from Brazil to China we’ve seen fantastic 3D printed surgical models that are used to greatly improve the chances of surgical success. These 3D printed models are usually based on CT or MRI scans to provide an accurate replica of the particular medical issue within the patient’s body, enabling surgeons to carefully prepare every step before cutting you open. While hitherto a time-consuming and complex 3D modeling process, researchers from MIT and Boston Children’s Hospital have just announced the completion of an efficient system for transforming MRI scans into 3D printable medical models within a matter of hours.
This new system will be completely described at the International Conference on Medical Image Computing and Computer Assisted Intervention in a few weeks from now, but it is already looking good. First author is graduate student Danielle Pace, studying electrical engineering and computer science at MIT. She spearheaded the development of the analysis software, while indispensable work was also done by Boston Children’s Hospital physicist Medhi Moghari, who developed procedures for optimizing scan precision. Cardiologist Andrew Powell is leading the clinical side of the study. Everything was funded by the Boston Children’s Hospital and Harvard Catalyst, which focused on clinical scientific innovations.
This new system is still under development – an evaluation study by seven cardiac surgeons is planned in the fall – but it is looking very good already. As project leader and professor of electrical engineering and computer science Polina Golland explained on the MIT website, this could make 3D printed surgical models common, rather than unique pieces. ‘Our collaborators are convinced that this will make a difference. The phrase I heard is that ‘surgeons see with their hands,’ that the perception is in the touch,’ she says.
To explain, MRI data consists of a series of cross sections of any 3D model – even human body parts. The result are a series of cross section photographs with lighter and darker regions being visible. Now the borders of these regions have always been difficult to precisely pinpoint for computer visions, and algorithms were usually not seen as reliable enough for generating precise models used by surgeons. While augmenting algorithms with generic models of, say, human hearts, could make them more precise, that does little when surgery on irregular hearts is necessary. In fact, it could make it worse.
So what use is MRI data when making 3D printable models? Well, designers have relied on these cross sections when manually making models, but that easily takes 10 hours or more. ‘They want to bring the kids in for scanning and spend probably a day or two doing planning of how exactly they’re going to operate,’ Golland says about surgical conventions. ‘If it takes another day just to process the images, it becomes unwieldy.’
But the solution developed by Pace and Golland takes a middle way. For each 3D model, a human expert is brought into identify boundaries in a few cross sections. That data is fed into the image algorithms, which takes over. The results were very good, even if the expert looks at about a ninth of the total area. Segmenting just 14 patches from a total of 200 cross sections was enough to get a 90 percent agreement rate. ‘I think that if somebody told me that I could segment the whole heart from eight slices out of 200, I would not have believed them. It was a surprise to us,’ Golland said. Remarkably, this method takes about an hour to generate amodel, with 3D printing taking just a few more.
While still under development, Golland believes that the algorithm’s performance might be improved by adding patches that run across cross sections. Alongside a couple of other innovations, this could be the key to accurate and quick 3D modeling and printing.
To test everything, a clinical study is set to take place, which will involve 10 patients who have already been treated in Boston. Their MRIs will be studied by seven surgeons, who will be given physical 3D prints, 3D models, and data and will use that to develop surgical plans. These, in turn, will be compared to the actual treatment methods used and will hopefully point out areas where this technique can benefit surgeons and improve success rates.
However, as one cardiac surgeon Sitaram Emani says, the prospects are good. ‘A 3-D model would indeed help. We have used this type of model in a few patients, and in fact performed ‘virtual surgery’ on the heart to simulate real conditions. Doing this really helped with the real surgery in terms of reducing the amount of time spent examining the heart and performing the repair,’ he says. ‘I think having this will also reduce the incidence of residual lesions — imperfections in repair — by allowing us to simulate and plan the size and shape of patches to be used. Ultimately, 3D-printed patches based upon the model will allow us to tailor prosthesis to patient.’ What’s more, these models make it far easier to explain surgeries to patients and families, as it gives them a better understanding of what’s coming. That alone would be a fantastic development.