THURJ Volume 15 | Issue 1

Abstract

There is an alarming increase in the population of ticks and tick-borne diseases (TBDs), with 475,000 cases reported annually, some of which are fatal. Due to limited training, healthcare providers and the public cannot always accurately identify ticks and their associated infections, leading to delayed diagnoses and treatments. Additionally, the prevalence rates of different disease-causing pathogens vary based on geographic locations. To facilitate the identification process and provide an expedited risk assessment of TBDs, a machine learning-based iOS application, DETICKT IT was created. The app features a ResNet50V2 (transfer learning) deep convolutional neural network (CNN) built in Python for combining real-time tick-species identification with a location-based tick-risk assessment by embedding the Centers for Disease Control and Prevention’s (CDC’s) spatiotemporal tick and pathogen surveillance statistics. With DETICKT IT, users can now receive an immediate and accurate analysis to determine whether they are at risk of contracting a certain TBD. The app is able to accurately identify the ten most common tick species in North and South America:American dog tick (Dermacentor variabilis, D. similis), Asian longhorned tick (Haemaphysalis longicornis), Brown dog tick (Rhipicephalus sanguineus), Eastern blacklegged tick (Ixodes scapularis), Western blacklegged tick (Ixodes pacificus), Groundhog tick (Ixodes cookei), Gulf Coast tick (Amblyomma maculatum), Lone star tick (Amblyomma americanum), Rocky Mountain wood tick (Dermacentor andersoni), Soft tick (Ornithodoros). The overall accuracy is 97%, with precision, recall, and F1 score metrics of 0.96, 0.97, and 0.96, respectively. This freely accessible app shows promise in assisting tick bite victims with their decision to seek further medical assistance, particularly those with underlying health conditions.