Datensatz

Zusammenfassung

Introduction: Magnetic Resonance Imaging (MRI) is increasingly used in radiotherapy planning due to the wide range of soft tissue contrast, which facilitates the differentiation between tumor and surrounding tissue. These contrasts can be tuned by a vast variety of MR sequence parameter sets (SPS), which directly affect image quality parameters like signal- (SNR) or contrast-to-noise ratio (CNR). Depending on the sequence and clinical objective, these SPS can include up to 30 individual parameters. The aim of this work is to develop a Software tool for the optimization of MRI sequences with regard to the applied SPS for SNR and CNR.
Materials & Methods: First, a model to predict the quality parameters (SNR/CNR) depending on the applied SPS was developed and trained. For this, training data sets were acquired at a 1.5 T MRI (Sola, Siemens) with a dedicated phantom with in-house fabricated anthropomorphic contrast inserts of different concentration of agarose and Ni-DTPA. For this, measurements with either fixed bandwidth or number of signal averages (NSA) were used to generate a total of 40 different SPS-combinations. For SNR determination the SNRdouble method is used [1]. Then a generalized additive model (GAM) is used for regression which is based on spline functions.
Results: According to the physical dependency, SNR is expected to be proportional to and . The measured SNR data was compared with the physical dependency and the GAM predictions (Figure 1). The GAM predicts the given data with two independent parameters with a mean absolute error (MAE) of 0.6, and 3.2 for bandwidth and averages respectively.
Summary: As a first step towards the development of an optimization tool for MR sequences, the GAM method showed a good agreement with the presented measurements. The next step is to develop and implement optimization methods with regards to SNR based GAM.