AI‑supported automated microscopy for malaria diagnosis: multicenter evaluation of the Noul miLab in Ethiopia and Ghana

miLab™ device achieves near-expert-level accuracy for both P. falciparum and P. vivax while outperforming routine health-center microscopy in sensitivity and speciation.

medRxiv, preprint

June 11, 2025

Dawil Hawaria, et al.

Abstract

This multicenter study evaluated the Noul miLab, an AI-supported automated microscopy system for malaria, in febrile patients from Ethiopia and Ghana. A total of 2,201 samples were tested with local microscopy, miLab, rapid diagnostic tests, and compared against expert microscopy and qPCR as gold standards. For P. falciparum, miLab showed sensitivity of 96.3–97.4% and specificity of 98.8% compared to qPCR at >200 parasites/µL. For P. vivax, sensitivity was 95.9–96.8% and specificity 97.8%. miLab significantly outperformed routine health-center microscopy in sensitivity and correctly identified species in >96% of P. falciparum and P. vivax mono-infections.

	Result
P. falciparum sensitivity vs expert microscopy	96.3% (335/348)
P. falciparum sensitivity vs qPCR (>200 parasites/µL)	97.4% (298/306)
P. vivax sensitivity vs expert microscopy	96.8% (399/412)
P. vivax sensitivity vs qPCR (>200 parasites/µL)	95.9% (419/437)
P. falciparum specificity vs qPCR	98.8% (1057/1070)
P. vivax specificity vs qPCR	97.8% (617/631)
P. falciparum species assignment accuracy in Ethiopia	99.3% (147/148)
P. vivax species assignment accuracy in Ethiopia	96.5% (304/315)

Table 1: Summary of miLab™ Clinical Validation Metrics

Key Highlights

miLab™ showed high diagnostic performance, with up to 97% sensitivity and 98% specificity versus qPCR.
It outperformed routine microscopy and achieved strong species-level accuracy, including 99.3% for P. falciparum.
Its fully automated workflow improved standardization and reduced operator variability across the diagnosis process.

Read Paper Download Brochure Request Demo