Abstract

Loiasis is a blood-borne filarial infection under consideration for inclusion in the WHO’s priority list of Neglected Tropical Diseases. In addition to its direct impact on human health, loiasis frequently occurs as a co-infection with other filarial diseases, including onchocerciasis (river blindness) and lymphatic filariasis (LF, elephantiasis). This overlapping geographic distribution seriously hinders mass drug administration (MDA) with ivermectin (IVM) for onchocerciasis and LF control, because individuals with Loa loa microfilarial (mf) loads exceeding 30,000 mf/mL are at high risk of developing severe adverse events, including coma and death, following IVM treatment. Therefore, safe MDA in loiasis co-endemic areas requires active detection and exclusion of high L. loa mf individuals from treatment, the so-called Test and Not Treat (TaNT) strategy. Automated, AI-driven diagnostics have shown strong potential to enhance TaNT implementation and fill a crucial care need for underserved populations, yet important performance limitations remain. To support AI research for the pressing health care needs of filarial diseases, we release this first-of-its-kind dataset to enable development of AI algorithms for detecting L. loa in whole blood samples. The dataset includes over 33,000 videos of whole blood in capillaries, in 4705 sessions (7 videos per session) from 1948 patients, captured by NTDscope portable imaging devices during four studies in 2023 and 2025 in Cameroon and Gabon. The videos contain a range of L. loa microfilariae counts and negative cases, as well as some cases with Mansonella perstans, another tropical filarial parasite. The metadata include expert microfilariae counts from Giemsa-stained blood films for each patient.