Diagnosis of Plasmodium infections using artificial intelligence techniques versus standard microscopy in a reference laboratory.

Abstract

Diagnosing malaria using standard techniques is time-consuming. With limited staffing in many laboratories, this may lead to delays in reporting. Innovative technologies are changing the diagnostic landscape and may help alleviate staffing shortages. The miLab MAL, an automated artificial intelligence-driven instrument was compared with standard microscopy at LabCorp reference laboratories. Four hundred eight samples submitted for parasitic examination were prepared with thick and thin smears and Noul's malaria platform miLab MAL, and evaluated for positivity, negativity, percent positivity, and species identification. Of 408 samples, 399 samples were manually negative, while 397 were negative by miLab MAL. Two samples initially classified as negative manually were found positive by miLab MAL. In all nine cases, Plasmodium falciparum was identified by both methods. Percentage parasitemia was higher in the manually calculated method, especially when >1%. miLab MAL was accurate in identifying the absence of Plasmodium falciparum and exhibited higher sensitivity than the manual method. All positive samples detected by microscopy were also identified with miLab MAL. All positive Plasmodium cases were correctly identified by miLab MAL. However, the number of positive samples was limited to only Plasmodium falciparum. Although parasitemia by the manual method was on average six times higher than with miLab MAL, this may be due to sampling variability. The findings show that miLab MAL can be used to screen out negative Plasmodium falciparum samples. Further studies assessing parasitemia between methods and identification of non-falciparum samples are necessary to assess the reliability of this new technology.