Medical-AI trial: identifying snake species with AI to aid snakebite treatment — a deep-learning system tested in emergency care (64-class classification) (NCT07658885)
A multicenter prospective observational study that develops and validates a deep-learning AI system to identify snake species and evaluates its clinical usefulness in real emergency snakebite care. Across 10 institutions in Zhejiang, China, it tests whether the system can accurately identify indigenous biting species and improve the accuracy and efficiency of clinician species judgment. 64-class classification plus venomous/non-venomous discrimination. 400 cases.
Trial overview (primary data)
- StatusEnrolling by invitation
- ConditionsSnake Bite
- InterventionsDIAGNOSTIC_TEST: SSRS system-assisted snake species identification
- SponsorSecond Affiliated Hospital, School of Medicine, Zhejiang University
- Target enrollment400 participants
- Period2026-06-20 〜 2026-12-31
Key points
- A multicenter prospective observational study developing and validating a deep-learning AI snake-species identification system, evaluated in real emergency snakebite care.
- Runs at 10 institutions in Zhejiang, China; tests whether it accurately identifies indigenous biting species and improves clinician accuracy and efficiency.
- Primary outcomes are 64-class (64-species) classification and venomous/non-venomous binary discrimination. 400 cases, by invitation.
- For snakebite, response and antivenom choice depend on venomous status and species, so accurate identification shapes the outcome.
- At the identification-performance and usefulness stage; not an establishment of clinical use or replacement of specialist judgment.
- Whether the snake is venomous changes the response, and for a venomous one the choice of antivenom shapes the outcome.
1In snakebite, species identification decides
In treating snakebite, quickly identifying which species of snake is crucial. Response differs greatly for a venomous versus a non-venomous snake, and for a venomous one, choosing the species-appropriate antivenom shapes the outcome. But the biting snake is often not seen, and many clinicians are unfamiliar with species identification, so accurate identification is not easy.
This study examines whether a system that identifies species from images with AI can help here.
2Identifying species with deep learning
Per the registry summary, this is a multicenter prospective observational study that develops and validates a deep-learning-based AI snake-species identification system and evaluates its clinical usefulness in real emergency snakebite care.
The main question is whether, in real clinical settings in Zhejiang Province, China, the system can accurately identify common indigenous biting species and improve the accuracy and diagnostic efficiency of species judgment for physicians at institutions of different levels.
It runs simultaneously at 10 institutions in Zhejiang, with primary outcomes of 64-class (64-species) classification and venomous/non-venomous binary discrimination. The size is 400 cases, conducted by invitation.
3Identifying the species decides the treatment
In snakebite, telling quickly which snake it was is critical. Following what each stage decides shows where AI fits.
- 1The bite occursOften nobody saw the snake
- 2Identify the speciesWhether it is venomous changes the response entirely
- 3Choose the antivenomFor a venomous species, the choice of serum shapes the outcome
- 4AI assistsEvaluated on 64-class identification and on venomous-versus-non-venomous determination
Many clinicians are not practised at telling species apart, so accurate identification is not easy. This is a multicentre prospective observational study across ten sites in Zhejiang, covering 400 cases. Snakebite counts globally among neglected tropical diseases, and AI supporting physicians where specialists are scarce could help with appropriate antivenom use and transfer decisions.
Why it matters
Snakebite is a neglected tropical disease where species identification and antivenom choice are tied to saving lives. Measuring across centers whether image-based AI can support clinicians in primary care or specialist-scarce areas is a useful reference for AI in the field (this study evaluates identification performance and usefulness, not clinical use).
FAQ
Why does identifying the snake species matter?
Does the AI replace clinician judgment?
Sources (primary)
Source: ClinicalTrials.gov (U.S. NIH/NLM, public domain). This site does not provide medical advice. Verify the latest and exact details with the official source. This site is not endorsed or certified by the NIH/NLM.
- ClinicalTrials.gov (study record, original)
- NCT ID: NCT07658885