An Integrated Shape-Texture Descriptor for Modeling Whole-Organism Phenotypes in Drug Screening

Schistosomiasis is a parasitic disease with global health and socioeconomic impacts. The World Health Organization (WHO) and National Institutes of Health (NIH) list it among diseases for which new treatments are urgently required. Drug discovery for Schistosomiasis typically involves whole-organism phenotypic screening. In such an approach, the parasites are exposed to different chemical compounds, and systemic phenotypic effects captured via microscopy (video or still images) are analyzed to identify promising molecules. Changes in parasite phenotypes tend to be multidimensional, involving changes in shape, appearance and behavior, and time-varying. In many image representation frameworks, shape and appearance are measured independently and their inter-correlation can be lost. In this paper, we propose an integrated shape-texture descriptor called the skeleton-constrained shortest band (SCSB) that extends the family of shape context descriptors well known in computer vision. We examine how SCSB can be used to measure temporally varying shape and appearance changes occurring as a consequence of chemical action and compare its performance with other members of the shape context family.