Quality control (QC) is an essential part of medical imaging workflow. Measurements assessing image validity are performed to guarantee that diagnoses are based on adequate information. Traditionally, quality assurance (QA) with imaging modalities using ionising radiation has been under strict regulation. On the other hand, modalities based on non-ionising radiation, such as magnetic resonance imaging (MRI), have only very limited amount of regulation, although MRI is the most prevalent method in diagnosing various diseases. There are several guidelines on proposed MRI QA protocols by international associations where applicable QC methods are presented. These methods are based on imaging a so-called phantom or test object featuring the measurement targets. Depending on the applied test and the used phantom, the scanning of the phantom protocol may take from a few minutes up to several hours. Although, the phantom-based methods can define the MRI hardware performance in many aspects, the relationship between the phantom measurements and the actual clinical image quality may remain vague. In this Thesis, streamlined methods for carrying out phantom based QA in a large radiological department were studied. An automatic image processing pipeline was built to improve data analysis and presentation. Furthermore, automatic error detection methods were compared with human observers. Additionally, new methods for QC measurements directly from clinical head images were developed. The methods were able to measure image quality changes quantitatively by directly analysing the appearance of the brain in the images. In addition to QC, the presented methods would enable comparison between different scanners and quantitative optimisation of imaging sequences.
|Publication status||Published - 2018|
|MoE publication type||G5 Doctoral dissertation (article)|
- magnetic resonance imaging
- image analysis
- quality assurance
- quality control