Preliminary Study Of The Personalized Patient-based Virtual Heart Modeling Approach And Its Application In Ventricular Tachycardia And Atrial Fibrillation

Cardiovascular diseases are the number one threat to human health,and their morbidity and mortality rates are increasing globally year by year and tend to be younger.Ventricular tachycardia(VT)and atrial fibrillation(AF)are common clinical cardiac diseases,and the intrinsic mechanisms underlying the development and maintenance of VT and AF have not been fully investigated,resulting in many shortcomings in clinical treatment strategies,including poor intraoperative success rates and high postoperative recurrence rates.Given the limitations of clinical and experimental studies,an increasing number of research groups internationally are using computer simulation modeling to investigate the mechanisms underlying the onset and maintenance of various cardiac diseases and to optimize clinical procedure protocols using this approach.The patient-based personalized virtual heart model simulation technology is one of the most cutting-edge researches in the field of cardiac modeling simulation,and its application scope not only includes assisting clinicians in risk prediction of various heart diseases but also can assist doctors in developing personalized procedure plans,showing great application potential and development space.The research work in this thesis focuses on building a framework and preliminary clinical validation of patient-based personalized cardiac modeling simulation technology,as follows.1、A preliminary general framework for patient-based personalized VT modeling simulation was built.At the cellular level,a single-cell action potential model of normal and infarcted cells in patients with VT was constructed;at the tissue level,a personalized 3D cardiac simulation model was constructed,including the construction of a personalized 3D anatomical model of the patient,the automatic generation of myocardial fiber rotation,the construction of clinically similar VT stimulation protocols,and the simulation of the electrical excitation conduction of the heart under VT.2、A modified Gaussian mixture model(MGMM)is proposed to fully automate the segmentation of the infarct region.The algorithm first performs initial segmentation of clinical low-resolution LGE-MRI images of VT patients using the conventional GMM algorithm and then performs noise pixel removal,automatic extraction of infarct regions,and discrimination of pseudo-infarct regions on the segmented tissues using the algorithm proposed in this thesis.The algorithm was validated on 60 LGE-MRI images from different hospitals and different acquisition devices.The segmentation results showed that the algorithm has the advantages of being fully automatic,high accuracy,and high reproducibility compared with the commonly used clinical methods(n SD method and full width at half maximum method).3、Two different infarct segmentation algorithms(the proposed MGMM algorithm and the threshold method)were used to segment infarct tissue on clinical LGE-MRI images from different devices(GE,Simens),and then a personalized VT model was simulated based on the segmentation results,and the accuracy of the models constructed by these two segmentation algorithms was compared to predict the location of VT lesions.The simulation results showed that the threshold method has the following drawbacks:(1)The simulation results of the threshold model are quite sensitive to the threshold size chosen by itself;(2)The threshold size corresponding to the threshold model that best matches the clinical outcome of the patient is not constant but varies dynamically from patient to patient.MGMM algorithm can accurately segment images from different devices without adjusting any parameters,and the simulation results of the model constructed based on this algorithm match well with clinical measurements,indirectly proving that the MGMM method has stronger generality and robustness compared to the traditional thresholding method.In addition,the distribution of infarct tissue in the model is critical in determining the accuracy of the model in predicting VT.4、The effects of five different grid spatial resolutions(645 μm,495 μm,420 μm,355 μm,320 μm)on the accuracy of VT obtained from simulations based on the patient personalized model were investigated.The results showed that the simulation model was able to achieve the best balance between simulation time consumption and accuracy of simulation results when the grid spatial resolution of the model was 420μm,and the simulation results provided a reference for future application of personalized cardiac simulation models to clinical ablation procedures to meet the time requirements for clinical application,i.e.,from obtaining patient image data to constructing the personalized model,completing the simulation and providing the results needed for clinicians should be less than 48 hours.5、A preliminary model of AF including the individual fibrosis distribution of the patient was constructed,and the ability of different catheter ablation strategies to terminate AF was simulated based on the constructed model.The simulation results showed that the ablation outcome is not dependent on the number of ablation lines,but rather on whether the reentries driving AF are ablated.