Publication Details
Abstract
Parasitic diseases represent a central health burden to many parts of the world, especially in the resource-limited setting. The increased rates of infected populations by drug-resistant strains of parasites indicate the essential demand of new interpretable techniques of parasite control. This paper presents a two-modality system which combines morphological feature analysis with modeling synthetic drug resistance. Based on applying more than 34,000 high-resolution, microscopic parasite and host cell images collected in eight different parasite and host cell classes, five most important morphology characteristics, i.e. area, perimeter, circularity, aspect, and intensity were extracted and analyzed statistically. An in-silico drug resistance profile of four antiparasitic drugs (Chloroquine, Amphotericin B, Suramin, and Metronidazole) was also established by estimating literatures-informed conclusions. Graphical representations, including the boxplot, heatmap, and pair plot, were used to identify some relationships between the structural characteristics and resistance patterns. Specific morphological profile was found in species such as Leishmania and Trypanosome as well as in increased simulated resisting levels. The innovation of the present work is that it has derived an explanatory, low-complexity model linking phenotypic image characterization with hypothetical drugs profiles that can be applied to parasite diagnostics--a method that has hitherto been uncharacteristic. The model is portable and affordable and provides the scalable solution to early resistance detection and morphological classification, which are most relevant in conditions when advanced molecular tools are not readily available.