Improved monitoring of silicone implants
TECHNICAL UNIVERSITY OF MUNICH
Corporate Communications Center
Improved monitoring of silicone implants
New algorithm for clinical MRI scanners makes breast examinations faster and more reliable
Magnetic resonance imaging (MRI) is a sensitive method for monitoring silicone implants. However, reliable implant examination can be challenging due to the difficulties of separating silicone and fat tissue in the images. An interdisciplinary research team at the Technical University of Munich (TUM) has now developed a new algorithm that improves the quality of MR images by depicting water, fat, and silicone simultaneously in a reliable and fully automated way.
Silicone implants are commonly used for breast reconstruction after mastectomy and for breast augmentation. Examining these implants at regular intervals helps to detect complications like implant rupture or implant associated anaplastic large cell lymphoma (BI-ALCL) early. MRI is the most sensitive available method for monitoring silicone implants. However, until now, it has been challenging to separate fat and silicone in MRI scans, as these materials generate a similar signal in MRI due to their similar frequencies under the main magnetic field of the MR scanner.
Improved MRI of silicone implants in the clinics
A research team led by Dimitrios Karampinos, Professor of Experimental Magnetic Resonance Imaging at TUM, has now developed a novel processing algorithm to solve the problem of reliably separating water, fat and silicone in MR images. It is based on a specialized data acquisition scheme encoding multiple chemical species previously proposed by his team. “From a technical viewpoint, it is challenging to obtain reliable information about water, fat, and silicone simultaneously with one scan. By collaborating closely with the radiologists at TUM, we not only solved this complex optimization problem from a mathematical perspective, but we developed a solution that can easily be implemented in common clinical MRI scanners and integrated in a clinical workflow”, says Prof. Karampinos.
The imaging process with the new method is simple and the data processing is completely automated. It is intended for breast imaging in all patients, with and without implants. The algorithm uses three main principles to solve the complex problem:
• Hierarchy: The algorithm processes the output from the MRI along a series of steps to create high-quality water, fat and silicone images. At each step, the algorithm decides how to proceed based on the available information, such as for example “implant or no implant”.
• Multi-resolution: High-resolution images depict a lot of detail, which can be helpful for diagnosis. However, high-resolution images are more susceptible to noise and could therefore make it difficult to separate water, fat and silicone. High-resolution images also need considerably more processing time. Hence, the algorithm starts with low-resolution images and increases resolution and complexity along the different steps when needed.
• Graph-cuts: To solve the complex problem of differentiating between several different chemical species, every value of the 3D MR image is encoded in a corresponding graph. However, one graph alone does not provide enough information to decide reliably if this value corresponds to fat, water or silicone. This is where the algorithm comes into play: Graphs are solved sequentially to build up on the previous information and to find the optimal solution – water, fat or silicone – that makes the most sense with respect to the constraints of the graph and within the context of the whole 3D image.
More robust, reliable and faster monitoring of silicone implants
Previous methods for MRI examinations are based on suppressing the other materials while imaging the material in question, for example silicone. However, this technique relies on several manual calibration steps, which can be prone to errors. While ruptures of implants can be imaged well with these available methods, slighter changes such as gel bleeding are more difficult to detect.
“The new method is fully automated and does not need any previous calibration and training for the operator. This makes it more robust and reliable than techniques that depend on the suppression of selected materials”, says Dr. Eva Maria Fallenberg, Adjunct Teaching Professor and senior physician at the Institute for diagnostic and interventional radiology at the University hospital Klinikum rechts der Isar. “So far, the new algorithm has shown reliable results even for different types of implants. What is more, acquiring all the information at once instead of one species after the other reduces total scanning time. This improves patient comfort, and allows us to examine more patients.”
After the first promising results, the new imaging process is now being evaluated with a larger cohort of people on clinical MRI scanners commonly used in hospitals. No extra equipment is necessary to adapt the new method. Hence, once the new method has been proven successful in a large patient cohort, an implementation in the large routine clinical setting might be expected. In the longer term, the researchers are also evaluating if the new method provides advantages for assessing non-implant tissue, as it might deliver additional useful information on measuring breast density and depicting breast calcifications.
J. K. Stelter, C. Boehm, S. Ruschke, K. Weiss, M. N. Diefenbach, M. Wu, T. Borde, G. Schmidt, M. R. Makowski, E. M. Fallenberg, D. C. Karampinos: Hierarchical multi-resolution graph-cuts for water-fat-silicone separation in breast MRI. IEEE Transactions on Medical Imaging (2022).
DOI: 10.1109/TMI.2022.3180302 https://ieeexplore.ieee.org/document/9788478
Prof. Dimitrios Karampinos and co-author Prof. Marcus R. Makowski are Principal Investigators at the Munich Institute of Biomedical Engineering (MIBE). MIBE is an Integrative Research Institute (IRI) within the Technical University of Munich (TUM) that fosters interdisciplinary cooperation and synergies between researchers from the broad field of Biomedical Engineering. At MIBE, researchers specializing in medicine, the natural sciences, and engineering join forces to develop new methods for preventing, diagnosing or treating diseases. The activities cover the entire development process – from the study of basic scientific principles through to their application in new medical devices, medicines and software. https://www.bioengineering.tum.de/en/
This work was supported in part by the European Research Council (ERC Starting Grant “ProFatMRI”). The authors also acknowledge research support from Philips Healthcare.
High resolution images:
Prof. Dr. Dimitrios Karampinos
Technical University of Munich
Professorship of Experimental Magnetic Resonance Imaging
The Technical University of Munich (TUM) is one of Europe’s leading research universities, with more than 600 professors, 48,000 students, and 11,000 academic and non-academic staff. Its focus areas are the engineering sciences, natural sciences, life sciences and medicine, combined with economic and social sciences. TUM acts as an entrepreneurial university that promotes talents and creates value for society. In that it profits from having strong partners in science and industry. It is represented worldwide with the TUM Asia campus in Singapore as well as offices in Beijing, Brussels, Mumbai, San Francisco, and São Paulo. Nobel Prize winners and inventors such as Rudolf Diesel, Carl von Linde, and Rudolf Mößbauer have done research at TUM. In 2006, 2012, and 2019 it won recognition as a German "Excellence University." In international rankings, TUM regularly places among the best universities in Germany.