In the field of stereology, the nucleator is a well-established manual method for estimating cell volume using measurements on a random cell transect through a fixed reference point of the cell. Here, we present an automized version of the nucleator that utilize automatic segmentation of the cell transect boundary. The segmentation process is supervised by an expert. In case the expert finds the segmentation to be satisfactory, then an estimate of the cell volume is calculated automatically on the basis of the whole cell transect. Otherwise, the supervising expert intervenes and the estimation of cell volume is performed using the classical nucleator. The resulting estimator is called the semi-automatic nucleator. We study the statistical properties of the semi-automatic nucleator and derive formulae for the bias and mean square error. Due to the automatic part of the estimation procedure the semi-automatic nucleator might have a small bias, though in most cases it will still be more efficient than the classical nucleator. We give procedures for estimating bias and mean square error from a pilot study. Use of the semi-automatic nucleator is illustrated in a study of somatostatin positive inhibitory interneurons which were genetically labeled with green fluorescent protein. An optical disector were used for the sampling procedure and as the reference point we used the centre of mass. From this study it was found that the number of cells needed to obtain e.g. a 5% precision of the estimate of mean cell volume is 150 and 189 for the semi-automatic and the classical nucleator, respectively. The semi-automatic nucleator is superior to the classical nucleator in the sense that the time spent analyzing one cell is shorter for the semi-automatic nucleator than for the classical nucleator.
Keywords: Computorized image analysis; Local stereology; Nucleator; Volume
Biography: Linda V. Hansen is a Ph.D. student at Centre for Stochastic Geometry and Advanced Bioimaging, Aarhus University, Denmark. Her thesis deals with topics within stereology, spatial statistics and stochastic geometry.