Advanced Image Processing in Cardiac Magnetic Resonance Imaging with Application in Myocardial Perfusion Quantification

Cardiac magnetic resonance imaging (CMRI) has been proven to be a valuable source of diagnostic information concerning heart health. One application, myocardial blood flow (MBF) quantification using first-pass contrast-enhanced myocardial perfusion, has aided the detection of coronary artery disease and provides an accurate evaluation of myocardial ischemia, an identifier of coronary artery stenosis. However, the image processing and analysis requires tedious user interaction, increasing the time and effort required to utilize it. In addition, it can introduce subjectivity and variability into the data analysis, which further limits the potential use of the modality. This dissertation presents several automated image processing algorithms to increase the accuracy, consistency, and efficiency of CMR image processing, and validates them on large, clinical datasets.First, an automated method is proposed to measure the arterial input function (AIF) from the left ventricle (LV), which is required for the accurate quantification of MBF. The proposed algorithm consists of several automated image processing steps including motion correction, intensity correction, detection of the LV, independent component analysis, and LV pixel thresholding to calculate the AIF signal. The method was validated in 270 clinical studies by comparing automated results to manual reference measurements using several quality metrics. Additionally, the MBF was calculated and compared in a subset of 21 clinical studies from healthy volunteers using the automated and manual AIF measurements. The proposed method successfully processed 99.63% of the image series. Manual and automatic AIF measurement showed strong agreement, and the automated method effectively selected bright LV pixels, excluded papillary muscles, and required much less processing time than the manual approach. No significant difference was found in MBF estimates between manually and automatically measured AIFs.Second, this dissertation presents an automated method for segmenting the myocardium from MBF maps, making segmental analysis faster and easier to achieve. The proposed method employs active contours for myocardial segmentation, and landmark detection for the anchoring of sector-wise analysis. These methods were validated in a group of 91 clinical perfusion studies against a manual reference standard. The proposed method processed 100% of the studies successfully and results agreed with the manual reference standard, both in terms of segmented area and measurements from sector-wise analysis.Together, these automated methods form a fully automatic MBF quantification pipeline for first-pass contrast-enhanced myocardial perfusion imaging. These advancements make the modality more readily available and applicable to a larger number of patients and centers throughout the field.Computer scienceMedical imagingCardiac Magnetic Resonance Imaging, Image Processing, Myocardial Perfusion Imaging, Quantification, SegmentationElectrical Engineering and Computer ScienceDegree Awarded: Ph.D. Electrical Engineering and Computer Science. The Catholic University of Americ