Quantitative analysis of myocardial infarction is important for the prognostic assessment and therapeutic strategy of patients with ischemic heart disease (IHD). 18F-fluorodeoxyglucose (FDG) PET is generally regarded as the gold standard for in vivo assessing myocardial viability. However, currently PET scans can only be interpreted by well trained radiologists because the morphological variation and location of infarct myocardium is hard to reflect visually on 2D slices or polar plot. In this paper, a 3D non-rigid coronary CTA (CCTA)-PET registration strategy is presented to quantitative analysis the extent of myocardial infarction. Whole heart is firstly segmented from both CCTA and PET/CT. Then the myocardium of left ventricle (LV) is separated from CCTA and PET/CT, respectively. After morphological post-processing, the myocardium image is integrated with the whole heart image both for CCTA and PET/CT to conduct the myocardial registration. A random forest classifier is trained to identify the infarct area of LV wall. The myocardial infarction analysis method proposed in this paper is compared with the cardiac tissue histological study. The result demonstrates good agreement with the TTC stain result in both infarct size and location, and suggests a potential value for clinic application in the prognosis of myocardial infarction.