Evaluation of the accuracy of smartphone applications with integrated artificial intelligence to identify species of hard ticks that commonly bite people in the United States

Abstract

Tick-borne diseases are on the rise in the United States. As tick species differ with regards to the suites of human pathogens they transmit, species identification of ticks found biting people may be helpful in assessing tick-borne disease risk. However, neither the public nor medical professionals are well-versed in tick identification. Artificial intelligence (AI) assisted object identification applications are available on smartphones but their accuracy for identifying ticks of different species, life stages, and engorgement levels is unknown. We evaluated 3 applications freely available on both iOS and Android platforms: 1 generalist object identification application (Google Lens) and 2 applications focused on nature identification (Seek by iNaturalist, Obsidentify). These applications were tested for their ability to identify ticks across five different smartphones and associated smartphone users. The evaluation included 5 species of human-biting hard ticks: Amblyomma americanum (Linnaeus), Amblyomma maculatum Koch, Dermacentor variabilis (Say), Ixodes scapularis Say, and Rhipicephalus sanguineus (Latreille). Across all species, life stages, and engorgement levels, the overall identification accuracy was low with 54% of specimens correctly identified to taxonomic Order (Ixodida [ticks]) and 16% accurately identified to species. Accuracy was highest for unfed adult females but generally poor for unfed immature ticks and blood-fed specimens. Our study revealed a high degree of variation in identification accuracy among tick species, life stages and engorgement levels. Currently, these tools have limited utility in clinical settings, due to their low accuracy at the clinically-relevant genus and species taxonomic levels. Future testing is needed as these applications continue to evolve.