Mosquito age grading from near infrared spectroscopy

Abstract

A child in Africa dies from malaria every 43 seconds. Because of the threat that malaria poses to human life, there have been renewed concerns recently for global elimination of the disease, the primary tools being vector control. As such measures are scaled up in areas with high malaria prevalence, there is a need to know the rate of transmission of the disease and the impact of vector control schemes in such areas. Two common approaches to attempt this are ovary dissection, which is tedious and expensive, and more recently Near Infrared Spectroscopy (NIR), which can be a complementary method to ovary dissection. In this thesis, we surveyed common pre-processing techniques and demonstrated the effect of five selected pre-processing techniques (multiplicative scatter correction, extended multiplicative scatter correction, standard normal variate, Savitzky Golay smoothing and gap segment derivative) on mosquito age estimation from NIR data. The selected pre-processing techniques were compared based on how well applying them improved mosquito age prediction using the root mean square error of prediction (RMSEP) and Q^2 values. Our results generally show that proper pre-processing improves mosquito age prediction. Scatter correction methods were also observed to perform better than techniques belonging to the spectral derivative category. Multiplicative scatter correction (MSC) gave better results compared to the other pre-processing techniques, while extended multiplicative scatter correction (EMSC) recorded the worst performance.