Towards reconstructing a volumetric data set from dual-isotope autoradiographic images

Volume based analysis of autoradiographic images is hampered because such images are primarily available as a series of 2D slices with no exact spatial slice-to-slice relation. To quantify the tracer uptake in extended brain regions requires a consistent volume of data. Digitized autoradiograms from a dual isotope rat brain study (18-F-FDG and 3-H-DPCPX) were available as a number of separate slices assembled in an image matrix of 8192×10240 pixels. Slices were extracted from the image matrix by an automated algorithm and stored as a raw volume of slices in 256×256 matrices. The stacks of slices from both isotopes were merged slice-wise into a combined stack, with two consecutive slices originating from an identical cut. The procedure to obtain a consistent volume relies on the similarity of neighboring cuts from the rat brain, and on a sufficient similarity, i.e. tracer uptake pattern, of the two autoradiographic slices deriving from an identical cut. Always a pair of neighboring slices from the merged stack are registered by applying appropriate shifts and rotations, while optimizing a similarity measure like mutual information. Optimization was performed according to the Simplex algorithm and resulted in a spatially aligned combined volume of slices, which was subsequently divided up into two volumes, one for each isotope.