Field Evaluation of a Device for Automated Malaria Microscopy (Autoscope Version 2)

Microscopy remains a key indicator in drug efficacy testing performed in the context of clinical trials for monitoring existing antimalarials or in the context of regulatory clinical trials for registration of new drugs. It is one of the main diagnostic methods for malaria diagnosis in general, as in an ideal setting it can provide low-cost accurate diagnosis, determine the density of parasites in the blood, and accurately differentiate between different malaria parasite species, characteristics vital to the implementation of global plans for drug efficacy monitoring. Malaria rapid tests (RDTs), while useful for case management, do not provide information on the parasite density nor the species differentiation necessary for research and drug efficacy assessment. Microscopy therefore retains key advantages over a number of newer technologies, but its reliability is severely impeded by dependence on high technical competence of the human operators as well as availability of high quality equipment and reagents. Recent studies have demonstrated the frequent poor specificity and sensitivity associated with manual microscopy diagnostics in operational conditions , , . Advances in digital microscopy performance and affordability have now opened the door to potentially significant improvements in the performance of malaria diagnostic microscopy, overcoming serious deficiencies in current drug efficacy assessment, and more broadly in malaria diagnosis and management. Intellectual Ventures Laboratory (IVL), in collaboration with Global Good Fund (GG), has developed an initial microscope prototype to support its research into dark field imaging of unstained malaria slides. The system consists of low cost electromechanical components for scanning a standard slide, an optical train with a high numerical aperture objective, and an image capture system. Captured images are analyzed with custom image analysis software developed at GG/IVL, using algorithms that are designed for aut