Identification of Plasmodium sporozoites using angle-resolved scattered light pulse shapes

Flow cytometry is widely used to identify and sort rare cells, but conventional approaches rely on fluorescent labels. Label-free discrimination of rare targets in complex, heterogeneous samples remains challenging. Here, we introduce a label-free strategy that combines multi-angle pulse shape flow cytometry (MAPS-FC), which exploits the full temporal structure of the signal pulses during cell or particles transit through the laser spot, with a deep learning-based analysis. As a biologically relevant test case, we focus on Plasmodium falciparum sporozoites, the infectious stage of the malaria parasite, which need to be isolated from mosquito salivary glands and are embedded in a heterogeneous mosquito-derived background. Using a GFP-expressing reporter line for validation, we trained a long short-term memory (LSTM) neural network to classify sporozoites solely from their scattering pulse shapes captured through MAPS-FC. Using training data from multiple days and setting a threshold on the classification confidence, the network can distinguish sporozoites from mosquito debris without fluorescent labels. This work establishes MAPS-FC with deep learning as a platform for label-free identification of rare parasites in complex biological samples.