Diagnostic performance of the miLab MAL AI-driven automated digital microscopy system versus standard microscopy for Plasmodium spp. detection

Microscopic detection of Plasmodium spp. using thin and thick blood smears remains the gold standard for diagnosing malaria, but its performance is highly dependent on stain quality and microscopist expertise. Recently, artificial intelligence (AI)-based technologies have enhanced laboratory performance in malaria diagnosis. This study aimed to evaluate the performance of miLab MAL (Noul, Inc., Republic of Korea), an AI-assisted microscopy system designed to detect Plasmodium species using thin blood smears. Over a 6-month period, 162 blood samples from patients with suspected malaria or undergoing post-treatment follow-up were prospectively analyzed using conventional diagnostic methods and compared with miLab MAL results. Conventional methods identified 83 positive and 79 negative samples, whereas miLab MAL identified 76 positive and 86 negative samples. Compared with loop-mediated isothermal amplification (LAMP) and/or thick blood smear, miLab MAL showed a sensitivity of 90.4%, a specificity of 98.7%, and a Cohen's kappa of 0.89. When compared with thin smear microscopy alone, sensitivity increased to 97.4%, with a specificity of 97.7% and a kappa value of 0.95. Regarding species identification, 11 P. falciparum were not correctly identified at the species level, while two P. ovale and one P. malariae were misidentified as P. vivax by the miLab MAL. Overall, the miLab MAL tended to underestimate parasitemia. This first evaluation of miLab MAL in France shows a promising performance for Plasmodium detection, although improvements are needed for species differentiation and parasitemia quantification. With further optimization, this approach may become a valuable tool for malaria diagnosis in resource-limited or high-burden settings.IMPORTANCEArtificial intelligence (AI)-assisted microscopy is a promising approach to improving the consistency and accuracy of malaria diagnosis, particularly in laboratories with limited microscopy expertise. In this study, we evaluated the performance of the miLab MAL, an AI-assisted digital microscopy system developed for the rapid detection and identification of Plasmodium species in clinical samples. The system demonstrated high sensitivity and specificity for Plasmodium detection, suggesting its potential value as a diagnostic support tool in routine laboratory settings. However, limitations were identified in species-level identification and parasite quantification.