• AI Health Checks

    • STI AI - The Science

    Clinically validated STI AI screening on any smartphone

    The Science behind

    Helfie's STI AI 

    Helfie iSpySTI AI turns a smartphone into a private, evidence-based STI screening tool.

    Developed with leading experts from Monash University and Alfred Health, and powered by data from 10,520 patients at the Melbourne Sexual Health Centre (MSHC), the AI delivers accurate, personalised STI risk assessments within seconds.

    Using a Bayesian network model, the system analyses user-reported symptoms, sexual health history, and guided visual inputs to estimate the likelihood of common STIs and genital conditions. Clinician-confirmed diagnoses from MSHC form the gold-standard dataset that trains and validates the model.

    The result is simple, affordable screening that supports early detection, reduces stigma, and improves access to sexual healthcare.

    Why this matters?

    STIs are widespread, often asymptomatic, and frequently missed without routine screening¹¹.

    Up to 70 percent of women and 50 percent of men with chlamydia have no symptoms, contributing to silent transmission and delayed treatment12. Rising national and global rates place increasing strain on health systems, with an estimated $16 billion annual cost in the United States alone13.

    Barriers such as stigma, privacy concerns, clinic availability, and long wait times continue to prevent early diagnosis11, 14. Digital tools are shown to improve engagement and access, especially for young people, LGBTQ+ individuals, and people in remote areas15.

    Helfie iSpySTI provides private, timely risk insights that help users understand their symptoms and seek care sooner.

    The science in your smartphone

    The STI AI model uses a Bayesian decision structure trained on real-world clinical data to classify 12 common STIs and related conditions. It processes:

    • Symptom type and duration
    • Anatomical location
    • Demographic and behavioural risk factors
    • Optional guided visual information

    Five-fold cross-validation and testing on unseen cases demonstrated high screening performance, with:

    • Sensitivity: 80–95 percent
    • Specificity: 38–76 percent
    • Accuracy: up to 97 percent depending on condition and gender.

    Independent Validation 


    Performance was benchmarked against clinician-confirmed diagnoses at MSHC.

    Performance metrics include¹: 

    • Gonorrhoea (men): 96.8% accuracy
    • Herpes (women): 95.3% accuracy
    • Pelvic inflammatory disease: 87.3% accuracy
    • Balanitis (men): 85.5% accuracy
    • Bacterial vaginosis (women): 81.6% accuracy

    Accuracy remained consistent across diverse ages, presentations, and skin tones, supporting broad, equitable use.