DKFZ and Max Planck Researchers Ensure Safety of Advanced MRI Systems Using Virtual Human Models
Algorithms Improve Patient-Specific MRI Testing
“MATLAB provides an environment where implementing mathematical problems can be achieved with little programming effort. The broad range of tools and the ease of use … let me write algorithms quickly and with high computational efficiency.”
Key Outcomes
- Computational modeling and simulation were faster because domain experts could focus on scientific problems instead of spending time with programming specifics
- Precise real-time SAR calculations using virtual human models laid the foundation for MRI studies that can advance scientific understanding of the human brain without compromising safety or regulatory compliance
- Seamless transfer and deployment of code from DKFZ to Max Planck facilitated easy collaboration and use of algorithms in high-performance computing settings
The workflow for specific absorption rate (SAR) calculations starts with a virtual human 3D model. A map of specific virtual observation points (VOP) then feeds matrices and compression algorithms.
The Division of Medical Physics Radiology at the German Cancer Research Center (DKFZ) in Heidelberg advances image-based diagnostic and therapeutic procedures. Scientists from DKFZ are collaborating with colleagues from the Department of Neurophysics at the Max Planck Institute for Human Cognitive and Brain Science (MPI CBS), which uses advanced magnetic resonance imaging (MRI) techniques to explore the brain.
While MRI radio waves can be used to create images, they also heat body tissue. To ensure patient safety, limits for this exposition—called the specific absorption rate (SAR)—are set at 4 watts per kilogram for the whole body and 10 watts per kilogram for any part averaging a volume of 10 grams. Older MRI machines use a single channel to send these radio waves, making SAR calculations straightforward. In contrast, modern MRI machines utilize multiple channels to enhance image fidelity. However, this multichannel approach complicates SAR calculations, as they now depend on a combination of the amplitudes and phases of signals from all channels. This involves complex math with matrices that represent the electric fields and tissue properties, derived from simulations of virtual human body models—also known as in silico medicine. These simulations create millions of data points, making real-time SAR monitoring during an MRI scan very difficult.
Developing efficient compression algorithms that can help simplify this data and speed up SAR calculation is an active area of research. At DKFZ, Dr. Stephan Orzada is using MATLAB® to develop compression matrices that enable faster data computation without compromising accuracy. MATLAB enabled Dr. Orzada to write algorithms quickly and with high computational efficiency without needing to optimize code. He used Parallel Computing Toolbox™ to speed up computations and Optimization Toolbox™ to develop compression algorithms.
Dr. Mikhail Kozlov, a scientist at the Max Planck Institute for Human Cognitive Brain Science, uses Parallel Computing Toolbox and MATLAB Parallel Server™ to solve DKFZ algorithms on modern supercomputers. Dr. Kozlov uses an ultra-high field, multichannel MRI scanner to understand human brain activity. His research greatly benefits from patient-specific SAR model calculations. He and his colleagues aim to use structural MRI scans to create a patient-specific model and then utilize this model the next day for safety calculations of functional MRI scans. This rapid turnaround is made possible by the advanced DKFZ algorithms and the scaling capabilities in MATLAB. This collaboration has resulted in a more efficient patient-specific SAR assessment process, enhancing MRI safety and imaging quality.
Related Resources
