Methodology And Application Research Of Thermometry Based On Magnetic Resonance Imaging

In practical clinical applications of tumor thermotherapies,an ideal thermotherapy should completely burn the tumor while collateral damaging to the surrounding healthy tissues is minimized.The resulting shape and size of the burnt tissues during the thermotherapy depend on the distribution of heat received by the tumor and the surround tissues.To enhance the efficiency of the thermotherapy treatment,visually monitoring and controlling of the temperature of the living tissue in real-time are crucial during the thermotherapy procedure and doing so may help the surgeons to decide on the fly when to end the thermotherapy at an appropriate timepoint.Magnetic resonance imaging(MRI)thermometry is a noninvasive and effective method based on the sensitivity of several MR parameters relevant to temperature and has progressively been used in clinical treatments for measuring temperature in living tissues.Clinical studies have demonstrated that many diseases may lead to dysfunction of thermal regulation in patients,and MRI thermometry has been used to explore the underlying brain mechanism.MRI thermometry is primarily characterized by proton resonance frequency(PRF)of water,T1 relaxation time and molecules diffusion coefficient(D).The research of this thesis primarily focuses on the investigation of several hot issues in methodology and application of MRI thermometry,and thereby has proposed effective solutions addressing these issues.Methodologically,we first used several optimal schemes to improve the accuracy and efficiency of data processing of PRF thermometry,and thus proposed a normalized working flow,which basically limited the thermometry error in a scope within±1~oC,whereas the accuracy of temperature measurement reported in the previous studies was about±2~oC.Subsequently,we investigated the efficacy of a series of strategies that may correct the main magnetic field drift when PRF thermometry method is employed.We examined the efficacy using different fitting methods,different reference phantoms varying the types of the component oil and locations where to put the reference phantom.We found it optimal for PRF thermometry when quadratic fitting correction was combined with vegetable oil as an internal reference.Second,to further improve the PRF thermometry,we proposed an automatic thermometry method based on multi-modal MRI data,which involved the use of several state-of-the-art approaches,including the referenceless PRF method,variable flip-angle technique and diffusion-weighted imaging(DWI).This proposed method effectively overcame multiple weaknesses found in PRF thermometry,such as motion,magnetic field drift and fat insensitivity.Finally,to improve the efficiency of the previous T1 thermometry,we developed a back propagation(BP)neural network model for T1 thermometry,which was trained to map the non-linear T1 thermometry relationship using temperature and T1-weighted images acquired at low temperatures.The BP model was then extended to predict temperatures corresponding to T1-weighted imaging data acquired at higher temperatures.Our results showed that this approach was highly useful with improved accuracy in measuring temperature.Applying our work to clinical studies,we also investigated the differences of four MR pulse sequences at a field strength of 7T that are based on PRF thermometry,and the experimental results of using phantoms and porcine brain tissues both suggested that the EPI-GRE(segmented gradient-echo echo planar imaging)sequence was the best among the four sequences.In addition,we investigated the effects of using different MRI parameters in terms of magnetic field strength,pixel resolution and b-value on DWI-based thermometry.Our results showed that b-values within the range from 300 to 800 s/mm~2 were more appropriate for DWI thermometry.We found that the accuracy of DWI thermometry at a mid-level of in-plane resolution of 1.8×1.8 mm~2 was relatively better than those measured at a low resolution of 2.5×2.5mm~2 or at a high resolution of 1.25×1.25 mm~2.Moreover,results obtained at 7T were more accurate and stable at b=1000 and 1200 s/mm~2 than those at 3T,especially when pixel resolution was below 2.5×2.5 mm~2.These findings provided a concrete value for reference in selecting the sequences or the sequence parameters in the studies of MRI thermometry at 7T in the future.Additionally,MRI thermometry can be used not only in the clinical treatment,but can also be applied to scientific research.We thereby creatively combined DWI thermometry with resting-state functional MRI,to explore the alterations of intraventricular brain temperature(BT),and also the interrelationship between intraventricular BT and spontaneous brain activities(SBAs)in schizophrenics and stroke patients,respectively.We found that the difference in the intraventricular BT in either patients with adolescent-onset schizophrenia or those with auditory verbal hallucinations(AVHs)was not statistically significant,when both were compared with matched controls.However,the schizophrenia patients displayed an uncoupling effect between BT and SBAs in several disease-related regions in the brain,and AVHs showed a significant effect of modulation on the relationship between BT and SBAs.Stroke patients with different hand outcomes displayed a frequency-specific difference between subgroups in the SBAs.No significant differences were found between stroke patients with partially paralyzed hands(PPH)and those with completely paralyzed hands(CPH)in the intraventricular BT,but CPH group showed a coupling effect between intraventricular BT and SBAs in several regions compared with PPH group and this effect was modulated by frequency.Finally,to facilitate the clinical applications of MRI thermometry,we also have implemented a toolbox with graphical user interface based on MATLAB(Matlab2012a;The Mathworks,Natick,MA,USA),which has integrated all the MRI-based thermometry methods thus developed in this thesis.