Meeting Banner
Abstract #2380

MR Pulse Sequence Design with Artificial Neural Networks

Nahal Geshnizjani1, Kenneth A. Loparo2, Dan Ma3, Mark A. Griswold4, 5

1Dept. of Electrical Engineering and Computer Science, Case Western Reserve University , Cleveland, OH, United States; 2Dept. of Electrical Engineering and Computer Science, Case Western Reserve University, Cleveland, OH, United States; 3Dept. of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States; 4Dept. of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States; 5Dept. of Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, United States

This work presents a framework for automatic Pulse sequence design. We used Artificial Neural Networks (ANN) with a novel sequential block structure to design a system to automatically construct MR pulse sequences. We are able to predict a pure but extended T2-weighted signal from a TrueFisp sequence. This method can be utilized in MR Fingerprinting by designing non-traditional randomized pulse sequences for quantitative imaging

Abstract PDF

Keywords

ability accuracy accurate accurately acknowledge applied approximate arbitrary architect architecture artificial assigned audience authors automatically available better biological biomedical block blocks blood bottom clinical complicated composed comprehensive computer concept conference configuration confirms consistent constant dashed degrees depicts dept design desired develop difficult door efficiency electrical engineering enlarged equation evaluated even evolution evolutions excitation excitations extended fingerprinting foundation function funding generated generation good grants hall handle highly hospitals includes individual infinite inputs inspired interconnected known lack like magnetization mark match medical minimizes model modifying muscle nearly nervous network networks neural neurons nontraditional novel offered opened operating optimized output outputs outside parallel part particular physics plays potential predict predicted prediction predictions previous process projects propose proposed pulse pure quantitative quick quickly radiology random randomized receives reserve role sample schematic science scientists selects sequential simulation solid solution solutions stage starts structure system systems target tissues traditional trained training transverse true types viability view western