The organ shape is an important factor that affects the propagation of modeling light in vivo. It can directly affect the reconstruction process of fluorescence molecular tomography (FMT). Manual segmentation of organs is complex and requires high-quality images, while automatic segmentation methods such as edge detection, region growing and active contour models have great limitations in dealing with complex medical images. We propose an automatic segmentation method based on active shape models (ASM) to accurately segment the images of mouse organs. Moreover, the light source reconstruction is realized based on L1 norm optimization. We carry out an experiment with the computed tomography (CT) data of a real mouse to explore the relation between organ image segmentation accuracy based on ASM and reconstruction accuracy. The experimental results show that the ASM method can replace manual segmentation without affecting the position reconstruction of light source, when compared with the popular Snake model-based segmentation algorithm.