In stool examinations, artificial intelligence (AI) has the capability to detect pathogens.
A study by Antti Suutala and Lundin in 2024, published in PLOS NTDs under a CC BY 4.0 license, revealed that AI can successfully identify parasite worm eggs in animal stool samples, including those from insect species that were previously overlooked by laboratory technicians. This breakthrough highlights the potential of AI in enhancing the diagnosis and treatment of parasite worm infections globally.
According to the World Health Organization, approximately 1.5 billion individuals, constituting nearly one-quarter of the population, are infected by parasitic worms residing in their gastrointestinal tracts. These infections can result in malnutrition and stunted mental development. However, due to a scarcity of trained professionals capable of identifying these infections, diagnosis and treatment often pose significant challenges.
Johan Lundin and his team at the Karolinska Institute in Sweden explored the utility of AI in addressing this issue. Their primary objective was to expand access to the diagnosis of parasitic insect infections to a larger population.
The researchers conducted a study involving around 1300 stool samples collected from college students in Kenya. These samples were processed by a local medical facility and then digitally scanned for analysis using a telescope. Subsequently, the AI examination was performed on the scanned images, which were transmitted wirelessly to the cloud for evaluation.
The AI system was trained to identify eggs from three different types of parasitic worms: the roundworm Ascaris lumbricoides, the animal whipworm Trichuris trichiura, and hookworms such as Ancylostoma duodenale and Necator americanus.
Comparative analysis demonstrated that the AI’s performance in detecting infections ranged from 76% to 96% accuracy compared to a skilled laboratory technician examining the samples manually.
Isaac Bogoch, a researcher at the University of Toronto who was not involved in the study, commended the research for its significant progress in identifying parasitic insect diseases, emphasizing the positive implications of this development.
Notably, the AI system exhibited the ability to identify infections even in cases where professional technicians had missed them, detecting 79 instances that had gone unnoticed. The AI’s error rate in falsely identifying infections in parasite-free samples was minimal, occurring in only 1 to 2% of cases.
Although the current processing time for each test using AI ranges from 20 to 35 days due to data upload delays, the researchers anticipate that with access to faster 5G networks and streamlined data transmission, the AI analysis could be completed within just 5 minutes. Moreover, cost estimates suggest that AI-based detection methods may prove to be more cost-effective than traditional human-centric approaches, as indicated by coauthor Nina Linder from Uppsala University in Sweden.
However, the researchers acknowledge the need to address ethical considerations, particularly regarding informed consent and data privacy, when deploying AI for health data analysis. Isaac Bogoch emphasizes the importance of maintaining transparency and ethical standards in utilizing this technology for medical purposes.
In conclusion, while the potential of AI in revolutionizing diagnostic processes is promising, it is essential to navigate the ethical implications and ensure responsible deployment within a robust regulatory framework.
