Investigation Of Several Image Processing Methods In Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) is widely used in clinical application. It is essential to post-process MRI data for clinical disease assessment and diagnosis. Thus some MRI post-processing methods for clinical applications are proposed and the main contents include:1. An algorithm for fast and accurate T2*mapping based on Auto-Regression on Linear Operations (ARLO) of dataR2*(Inverse of MR relexation time T2*, R2*=1/T2*) has strong linear correlation with iron concentration in biological tissue. The use of mono-exponential fitting of the MR signal decay to obtain the transverse relaxation time (T2or T2*) has been central to the development of many quantitative MR methods for mapping tissue properties. Currently, the non-linear least squares based Levenberg-Marquardt (LM) and the linear Log-Linear (LL) are the most commonly used methods for mono-exponential fitting. The iterative LM algorithm is generally regarded as more accurate but computationally more expensive than the non-iterative LL algorithm which is fast but known to be sensitive to noise and may not work well for short T2*in iron overload diseases. So we proposed a novel fast and accurate method for calculating T2called Auto Regression on Linear Operations (ARLO) and compare it with the conventional LM and LL methods using simulated and in vivo data (the iron overloaded livers (n=15) and hearts (n=1) at1.5T, as well as in healthy brains (n=2) at3T). In simulations, ARLO consistently delivered accuracy similar to LM and significantly superior to LL. In vivo mapping of T2*values, ARLO showed excellent agreement with LM, while LL showed only limited agreements with ARLO and LM. Compared with LM and LL in processing speed, ARLO was125and8times faster.2. Improved method for MRI left ventricle automated segmentationIt is known from priori LV segmentation on short axis4D Cardiac Cine MRI investigations that basal ventricle segmentation is difficult due to its complex anatomy compared to mid-ventricle and apical-ventricle. The purpose of this project was to develop a novel method for MRI left ventricle aumated segmentation. The improved method was able to automatically determine basal section and accurately segments Left Ventricular basal slices. The proposed segmentation algorithm was implemented as a plug-in in Osirix (an open source code DICOM viewer) and cine cardiac MRI data from30subjects were analyzed. Automated LV segmentation algorithm with this method was used to measure left basal ventricle volumes, left ventricle volumes and ejection fraction, and compared with expert manual tracing. Results showed that the success rates of identifying and correcting erroneous effusion at basal ventricle slices were100.0%and89.9%respectively. Automated segmentation results were well agreed with manual tracing measurements:the corrected left basal ventricle slice volumes were well agreed with manual tracing measurement (R2=0.957and0.904for end-diastole and end-systole respectively). LV End-Diastolic volume, LV End-Systolic volume and LV Ejection fraction calculations with this automated segmentation method were highly correlated with manual tracing measurement (R2=0.978,0.952and0.946respectively).3. Automated MR perforator flap angiography reporting systemPerforator flaps have been widely used in plastic surgery like breast reconstruction. But MR Perforator flap angiography reporting requires meticulous measurements of co-ordinates and diameters of perforator flap artery (PFA) and estimation of flap volumes. This process is time consuming and prone to manual data entry errors. We have developed a computer reporting system that automatically generates a report of PFA coordinates, diameters, MIPs and flap volumes. The reporting software was implemented as a plugin in Osirix DICOM viewer. Accuracy and speed of the Osirix reporting on PFA (coordinates, diameters, intramuscular course length, fat volume) was evaluated by comparing with measurements in the manual report (n=76) as well as operative notes and surgical feedback when available (n=46). In Results, Perforator coordinate locations were within0.5cm in50%of the cases and within0.9cm in100%of the cases. The fat volume reported using automated method was accurate to50grams in90%of the cases. The errors in vessel diameter measurements were less than1cm. The mean total time for creating each report using the traditional method was156minutes compared to77.5minutes using the Osirix system with a mean difference of78.5minutes (p<0.0001, CI57.37-99.63).