An assessment of methods for aligning two-dimensional microscope sections to create image volumes

This study assessed five different methods for aligning microscope images of Nissl-stained sections of mouse brain to form three-dimensional image volumes. Methods exploiting both image content and information from un-sectioned tissue were investigated. The accuracy of reconstruction was estimated using fiducials with known physical properties, demonstrating that methods exploiting tissue content produced distorted image volumes while a method using artificial fiducials produced the most accurate and unbiased alignment. Methodological issues relating to methods of volume reconstruction are discussed and it is recommended that methods using information from un-sectioned tissue be used wherever possible.