| Περίληψη: | Magnetic Resonance Imaging (MRI) has a crucial role in disease management, further allowing the identification of non-invasive imaging biomarkers for disease detection, diagnosis and monitoring response to therapy. Towards this direction, quality assurance (QA) of MRI scanners is a prerequisite for ensuring optimal imaging quality in the clinical setting and for providing standardized approach for the identification of imaging biomarkers (Osadebey, et al., 2017), (Osadebey, et al., 2018).
The current thesis is focused on evaluating image quality of three clinical MRI scanners in terms of a phantom-based study, for a time period of three years. A standardized QA protocol was adopted (Ihalainen, et al., 2011), (Chen, et al., 2004), using the American College of Radiology (ACR) ACR accreditation phantom (The American College of Radiology, 2005), (AAPM, 2010). According to the standardized image quality evaluation methodology, both quantitative and qualitative image quality indices have been utilized.
Quantitative image quality indices include: geometric accuracy, slice thickness accuracy, slice position accuracy, image intensity uniformity, percent-signal ghosting, homogeneity and signal-to-noise-Ratio (SNR). The effect of slice thickness and receiver bandwidth in SNR is also investigated. Furthermore, the feasibility of first order statistical image features acting as adjunct quantitative image noise indices, in the frame of slice thickness and receiver bandwidth on SNR, is also investigated. The basic hypothesis underlying the use of first order statistical features is mainly based on the recent use of texture features as clinical image quality surrogates in brain MR images (Osadebey, et al., 2017), (Osadebey, et al., 2018).
Qualitative image quality indices include low-contrast detectability and high contrast spatial resolution. Inter-observer variability in qualitatively assessed indices is also considered.
For each MRI system, compliance to national and international image quality guidelines is discussed. Long term performance reproducibility (system stability) is assessed in terms of standard deviations and coefficient of variation analysis of the quantitative and qualitative image quality indices.
Results of the current study suggest the feasibility of the adopted ACR phantom-based QA protocol in monitoring short- and long-term MR systems performance. Extraction of image texture descriptors in combination to pattern classification schemes (machine learning) is expected to contribute to the development of automated QC procedures, with increased consistency in estimation of image quality indices, towards QA optimization in MRI.
|
|---|
