3-D Rat Brain Phantom for High-Resolution Molecular Imaging

With the steadily improving resolution of novel small-animal single photon emission computed tomography (SPECT) and positron emission tomography devices, highly detailed phantoms are required for testing and optimizing these systems. We present a three-dimensional (3-D) digital and physical phantom pair to represent, e.g., cerebral blood flow, glucose metabolism, or neuroreceptor binding in small regions of the rat brain. The anatomical structures are based on digital photographs of the uncut part of a rat brain cryosection block. The photographs have been segmented into ventricles and gray and white matter and have been stacked afterwards. In the resulting voxelized digital phantom, tracer concentration in gray and white matter can be scaled independently. This is of relevance since, e.g., cerebral blood flow or metabolism are much higher in gray than in white matter. The physical phantom is based on the digital phantom and has been manufactured out of hardened polymer using rapid prototyping, a process in which complicated 3-D objects can be built up layer by layer. X-ray computed tomography and high-resolution SPECT images of the physical phantom are compared with the digital phantom. The detailed physical phantom can be filled bubble-free. Excellent correspondence is shown between details in the digital and physical phantom. Therefore, this newly developed brain phantom will enable the optimization of high-resolution imaging for recovery of complex shaped molecular distributions.

I. Introduction

Small laboratory animals such as rats and mice are the most widely used animals for experimental studies in neuroscience. In these animals, regional molecular aspects of behavior or disease can be studied with the aid of a wide range of in-vitro methods, including autoradiography and immunohistochemistry. These procedures require sacrificing the animal after which tissue is sliced, which can be very labor intensive. Another drawback of these in-vitro studies is that repeated studies within the same groups of animals cannot be performed, which often results in the requirement to sacrifice relatively large numbers of animals to reliably quantify dynamic processes. Furthermore, obtaining three-dimensional (3-D) digital images from two-dimensional (2-D) images of tissue slices is extremely tedious work. Stacking of slice-images requires advanced image processing to deal with tissue cracks and other tissue deformations that often result from cryosectioning. Despite progress made in image processing for combining histological or immunohistochemical slices [1]–[3], recovering 3-D digital images from slice images cannot be performed automatically and remains extremely labor intensive.