A Fast Binary Image Segmentation Algorithm and its Application to in Situ Hybridization Data

High dimensional and high-resolution gene expression data generated by the in situ hybridization (ISH) technique provides biologists a powerful tool to study gene functions. Nevertheless, a major challenge in analyzing such data is how to efficiently retrieve genes showing certain spatial expression patterns and/or genes showing similar expression patterns as the query gene. The development of a fast image segmentation algorithm and an effective statistical scoring method is useful to address the above question. In this paper we propose a fast binary image segmentation algorithm, which can efficiently detect the spatially restricted expression regions of a gene and/or co-expression regions between genes. A naive permutation approach is applied to assess the statistical significance of the detected regions. Our method is expected to drive ISH data analysis toward a more systematic and objective direction.