Preview

PDF, English Download (2MB) | Terms of use

Abstract

This thesis sets out to investigate the usefulness of diffusion imaging and diffusion kurtosis imaging for characterization of suspicious breast and ovarian lesions. The work was divided into three parts. To answer the research questions, the experiments with the phantom were conducted, and the analysis of the clinical data from retrospective and prospective study was done. The first part of this thesis was devoted to the experiments with the phantom designed for the breast coil. The aim was to assess the repeatability and reproducibility of quantitative measurements of apparent diffusion coefficient, which is the most commonly used quantitative parameter in clinical diffusion imaging of breast. The ability to control the spatial and temporal uniformity of apparent diffusion coefficient was demonstrated with the semi-automated quality control procedure using the self-written MATLAB script. The interesting feature of the breast phantom is the built-in thermometer which enables the temperature correction of apparent diffusion coefficient values. The readout-segmented and single-shot echo-planar imaging sequences with various setups of diffusion sensitizing gradient and parallel imaging acceleration factor were used. It was shown that this simple isotropic phantom can be used to monitor the variability between left and right side of the breast coil and to detect the differences in measurements with various protocol setup parameters. Among the readout-segmented echo-planar sequences, the highest spatial nonuniformity was found for bipolar gradient sequences, with the variability more prominent in the left side of the coil. Similarly, the bipolar single-shot echo-planar sequences showed higher spatial nonuniformity than the monopolar one, however the bigger variability was observed in the right part of the coil. Moreover, the spatial nonuniformity of apparent diffusion coefficient values varied with the parallel imaging acceleration factor. In summary, depending on the sequence parameters, significant inhomogeneities and right-left differences were observed, which ultimately could not be explained within the scope of this thesis. Therefore, due to the relatively high deviation of ADC values, reaching a maximum value of coefficient of variation of 8.69%, further investigation of the origin of these differences seems urgently warranted, particularly regarding discrepancies between the phantoms and patient measurements concerning relaxation times, presence of adipose tissue, and varying diffusion coefficients. As a conclusion, by incorporating the regular quality control procedure into the clinical routine, the variations in measurements can be explored, which could help to ensure the robustness of the measurements of qualitative diffusion parameters. Using this simple breast phantom, the reliability of measured apparent diffusion coefficient values between the coil sides, measurements days, and acquisition among one measurement slot can be estimated. This is especially relevant if the range of the values of apparent diffusion coefficient for benign and malignant lesion overlap, like it is in the case of breast lesions. The phantoms could potentially serve as suitable tools for selecting appropriate measurement parameters to enable more homogeneous diffusion parameters. In the second part, the influence of residual fat-related signal on quantitative parameters in diffusion kurtosis imaging of breast was investigated. The not fully suppressed signal from adipose tissue may corrupt the signal in the lesion area and distort the quantitative parameters. Therefore the research aim was to develop various alternative fitting models accounting for possible corruption with signal from fatty tissue. A data-set comprises the diffusion images of suspicious mammography findings taken in two study centers. The proposed modified fitting models performed better than the standard approaches. In the analysis by study site, the incorporation of fat correction terms improved the results in the first cohort of patients. In the second cohort, no improvement with the lack of adverse effect was found. This suggests that the proposed modified evaluation methods may potentially be applied to reduce the negative impact of the contamination of the signal in breast lesion with unsuccessfully suppressed signal from adipose tissue in quantitative diffusion kurtosis imaging. Finally, the last part of this thesis studies the application of diffusion kurtosis imaging in differentiation between benign and malignant ovarian lesions. Moreover, as the exclusion of malignancy is the main target of this diagnostic imaging procedure, the inconclusive findings need further verification through surgery and histopathological examination for definitive diagnosis. However, in the final diagnosis, the benign tumors outnumber malignant ones. Therefore the improvement of imaging techniques may avoid the unnecessary invasive management of benign adnexal masses. Although many studies showed that diffusion kurtosis imaging has potential for characterization of lesion in various parts of the body, it remained almost unexplored, to date of writing this thesis, for differentiation between benign and malignant ovarian lesions. Therefore the aim of this study was to investigate the reliability of quantitative kurtosis parameters for characterization of sonographically indeterminate ovarian lesions. Receiver operating characteristic analysis showed, that the chosen threshold yielded maximum sensitivity, that is all malignancies could be correctly diagnosed, and specificity giving an accurate classification of every second benign lesion. At the same time, the half of the patients could be accurately classified. This suggests that diffusion kurtosis imaging may have potential for reducing the number of unnecessary oophorectomies.