Automated Alignment of 2D Serial Images for 3D Reconstruction of Biological Structures

To characterize the complex genetics of mammalian brain architecture, researchers have long used mouse models that mimic specific human maladies to decipher the causes of the disease and to test new drugs and treatments. One of growing interests is to use imaging and computer vision methods for so-called neuroinformatics and neurocartography. In this paper, we propose a novel knowledge-based approach for the alignment of 2D serial images of the mouse brain, thereby allowing 3D reconstruction of functional or anatomical cerebral structures in a mouse brain by forming 3D stacks of 2D serial histological sections. In our implementation, heuristic rules supplied by human experts are implemented by use of fuzzy logics. Our results suggest that accurate alignment can be achieved in an automated manner despite the input images are not completely identical in structure.