Multimodal Fusion-based Method For The In Vivo Assessment Of The Heart

Myocardial Infarction(MI)is one of the most major fatal diseases worldwide.Myocardial infarction may lead to permanent loss of myocardial cells and the formation of scar tissue,which,once formed,will remain for a lifetime and lead to further diseases such as arrhythmia,heart failure or sudden cardiac death.The current treatment adopted after myocardial infarction is mainly a drug therapy.Such a therapy can improve cardiac function,but it is unable to regenerate the lost myocardial cells or restore a heart to its pre-morbid exercise tolerance,and therefore the long-term prognosis of a patient with myocardial infarction is unsatisfactory.Over the last 50 years,scientists and cardiovascular physicians have been trying to develop new therapies to improve the condition of damaged hearts.Accurate assessment of cardiac structure,cardiac repair progress and infarct size after myocardial infarction occurs is essential for the development of new therapies.Current imaging techniques for post-myocardial infarction assessment primarily involve single-modal assessment.Single-modal cardiac imaging has its own advantages and disadvantages in terms of sensitivity,availability,reproducibility,and feasibility,but fails to satisfy all the requirements for developing new therapeutic options.In this paper,we proposed a comprehensive cardiac noninvasive assessment method based on multimodal image fusion.The main work of this paper is as follows.(1)With regard to the assessment of left ventricular remodeling after myocardial infarction,an automatic calculation method of a three-dimensional spherical index based on coronary computed tomography angiography(CTA)images is proposed.The three-dimensional spherical index of the left ventricle is a powerful predictor for left ventricular remodeling within 6 months after acute myocardial infarction(AMI).Firstly,CTA images with pathological section staining data were collected from 20 small porcine myocardial infarction models,and the marker maps of the left ventricular myocardium were obtained by manually outlining the inner and outer walls of the left ventricular myocardium in the CTA images layer-by-layer under the guidance of experts,and the atlas of the left ventricular myocardium in CTA images was established.On this basis,the hybrid model growth algorithm was used to realize automatic segmentation of the left ventricular myocardial layer of the CTA images.The hybrid model growth algorithm first uses a multi-atlas method to perform the initial segmentation of the myocardial layer,then uses a region growth method to merge adjacent pixels whose gray values meet certain conditions into the segmented region,and finally uses an active shape model to correct the segmented region after growth.Based on the obtained myocardial automatic segmentation results,the three-dimensional spatial correction of the CTA images was performed,and the automatic extraction of the left ventricular cavity and the automatic selection of the long axis of the left ventricle were completed,and finally the automatic calculation of the three-dimensional spherical index was completed according to the definition.In this paper,we analyzed the relationship between the CTA image-based 3D spherical index and the infarct size resulting from pathological staining,and found that the two were statistically correlated.(2)With regard to the observational study of the cardiac repair process after myocardial infarction,an in vivo assessment method of neovascularization based on cardiac multimodal image fusion was proposed.Neovascularization is one of the key processes of cardiac repair after myocardial infarction.Currently,imaging studies of neovascularization are mainly focused on the validation of the targeting and effectiveness of radiolabeled probes,with few studies combining neovascularization imaging with established clinical imaging for a comprehensive observation of the repair process after myocardial infarction.Meanwhile,existing imaging commonly focuses on two-dimensional-based analysis,which may not provide a comprehensive and in-depth picture of the regression of myocardial infarction.In this paper,a combination of cardiac anatomy and information on myocardial perfusion,glucose metabolism,and neovascularization is integrated in a single image through image alignment fusion technology to achieve an image assessment with comprehensive information,laying the foundation for subsequent exploration of the mechanisms of neovascularization.In constructing the multimodal fusion assessment method,the anatomical structure of the heart was provided by the CTA images,and myocardial perfusion information was provided by 99 m Tc-methoxyisobutylisocyanine(99m Tc-sestamibi,99 m TcMIBI)single-photon emission computed tomography(SPECT)imaging,myocardial glucose metabolism information was provided by 2-deoxy-2-[18F]fluoro-D-glucose(18FFDG)positron emission computed tomography(PET)imaging,and neovascularization was provided by 68Ga-1,4,7-triazacyclonane-1,4,7-triacetic acid-(Arg-Gly-Asp)2(68GaNOTA-PRGD2)PET imaging.A contour/region-based registration method was adopted to achieve the alignment procedure,which was divided into two stages: a coarse alignment stage and a fine alignment stage.The target image for coarse alignment was a CTA image and the target image for fine alignment was a nuclide functional image.Both the coarse and fine alignment use the overall heart contour as an additional constraint to ensure the accuracy of the alignment.Finally,the cardiac multimodal fusion results were visually presented in a3 D space.It was demonstrated that the 3D visualization results of multimodal fusion could visually reflect the relationship between neovascularization and myocardial metabolism and perfusion.From the results in this paper,it was observed that the neovascularization signal appeared in the overlapping zone of severe hypoperfusion and hypometabolism,directly below the coronary artery obstruction.This phenomenon suggests that neovascularization may be derived from congenital,underdeveloped vessels.This observation provides 3D imaging support for the target area of therapeutic neovascularization.(3)To address the problem of accurate quantitative assessment of myocardial area size in cardiac infarction,a 3D quantitative assessment method of myocardial infarction based on cardiac multimodal image fusion was proposed.In the process of clinical treatment,no matter for the long-term management of patients with myocardial infarction or for the study of experimental cardiac patch preparation,personalized and precise quantitative assessment methods of myocardial infarction are necessary.On the one hand,clinical follow-up and treatment planning require an intuitive and easily interpretable assessment report;and on the other hand,the preparation of cardiac patches requires a clear definition of the size and location of the infarcted myocardial region and the criminal vessels corresponding to the infarcted region.To achieve this goal,in this paper,a quantitative analysis method of myocardial infarction based on multimodal image fusion is proposed.Firstly,the spatial alignment of cardiac anatomical structure and functional information is achieved through automatic segmentation and alignment.Then,based on the 3D texture features of the aligned images,a random forest classifier is used to achieve myocardial classification and establish a mapping relationship between non-invasive in vivo imaging and pathological gold standard.The accuracy of the multimodal myocardial infarction quantitative analysis method was first validated using a small porcine myocardial infarction model.The experimental results showed that the proposed method in this paper is in good agreement with pathological staining results.The proposed method was then applied to the evaluation of patients with old infarcts,and the obtained 3D fusion-based analysis results showed both the infarcted and decompensated regions and depicted the distribution of coronary artery main branch vessels,reflecting the relationship between myocardial damaged regions and coronary artery lesion segments in the 3D space.Finally,in this paper,a patient-specific cardiac patch was constructed based on the precise localization of quantitative cardiac infarct regions.It has been demonstrated that the 3D fusion quantitative assessment method can more accurately and intuitively assess the infarct size and show the relationship between the infarct area and the coronary artery,internal mammary artery,and ribs in the 3D space,and can be used to guide the design of cardiac patches and the planning of patch implantation repair surgery.