 |
    |
Phenotype Detection in Morphological Mutant Mice Using Deformation Features
Sharmili Roy1, Xi Liang2, Asanobu Kitamoto3, Masaru Tamura4, Toshihiko Shiroishi4, and Michael S. Brown1
1School of Computing, National University of Singapore, Singapore
sharmili@comp.nus.edu.sg
2National ICT Australia (NICTA), Australia
3National ICT Australia (NICTA), Australia
4National Institute of Genetics, Japan
Abstract. Large-scale global efforts are underway to knockout each of the approximately 25,000 mouse genes and interpret their roles in shaping the mammalian embryo. Given the tremendous amount of data generated by imaging mutated prenatal mice, high-throughput image analysis systems are inevitable to characterize mammalian development and diseases. Current state-of-the-art computational systems offer only differential volumetric analysis of pre-defined anatomical structures between various gene-knockout mice strains. For subtle anatomical phenotypes, embryo phenotyping still relies on the laborious histological techniques that are clearly unsuitable in such big data environment. This paper presents a system that automatically detects known phenotypes and assists in discovering novel phenotypes in 
CT images of mutant mice. Deformation features obtained from non-linear registration of mutant embryo to a normal consensus average image are extracted and analyzed to compute phenotypic and candidate phenotypic areas. The presented system is evaluated using C57BL/10 embryo images. All cases of ventricular septum defect and polydactyly, well-known to be present in this strain, are successfully detected. The system predicts potential phenotypic areas in the liver that are under active histological evaluation for possible phenotype of this mouse line.
LNCS 8151, p. 437 ff.
lncs@springer.com
© Springer-Verlag Berlin Heidelberg 2013