Robust Neuron Counting Based on Fusion of Shape Map and Multi-cue Learning
Automatic counting of neurons in fluorescently stained microscopic images is increasingly important for brain research when big imagery data sets are becoming a norm and will be more so in the future. In this paper, we present an automatic learning-based method for effective detection and counting of neurons with stained nuclei. A shape map that reflects the boosted edge and shape information is generated and a learning problem is formulated to detect the centers of stained nuclei. The method combines multiple cues of edge gradient, shape, and texture during shape map generation, feature extraction and final count determination. The proposed algorithm consistently delivers robust count ratios and precision rates on neurons in mouse and rat brain images that are shown to be better than alternative unsupervised and supervised counting methods.
KeywordsNeuron counting Machine learning Shape map Microscopic neuronal image Nuclei staining
We thank Dr. Dragan Maric for providing the image for Bioimage Informatics Conference 2015 Nucleus Counting Challenge. The work was partially supported by NIH NIMH R15 MH099569 (Zhou) and R21 NS094091 from NIH and a Seed Grant from the Brain Research Foundation (Ye).
- 3.Zhou, J., Peng, H.: Counting cells in 3D confocal images based on discriminative models. In: ACM Conference on Bioinformatics, Computational Biology and Biomedicine (ACM BCB) (2011)Google Scholar
- 5.Sobel, I., Feldman, G.: A 3×3 isotropic gradient operator for image processing. In: The Stanford Artificial Intelligence Project (SAIL) (1968)Google Scholar