A smartphone-based AI platform, CaptureTumor (CaT), was evaluated for early detection of ocular surface malignancies in a nationwide clinical trial. The system identified previously undiagnosed malignancies.
Background
Ocular surface malignancies are rare but can lead to significant morbidity if diagnosed late. Traditional screening methods often result in misdiagnosis as benign lesions, delaying necessary treatment. The integration of smartphone technology and AI may enhance early detection.
Data Highlights
Metric
Value
AUC (smartphone testing)
0.905
AUC (real-world screening)
0.977
Sensitivity
89.3%
Specificity
95.9%
Histopathologically confirmed malignancies
20
Key Findings
CaptureTumor (CaT) achieved an AUC of 0.977 in real-world screening.
The system identified 20 histopathologically confirmed malignancies, including 14 basal cell carcinomas and 6 malignant melanomas.
Nineteen of the identified malignancies were previously undiagnosed.
Patients identified through the app required fewer referrals to access specialist care compared to historical averages.
The app provided real-time guidance for optimal image capture to enhance diagnostic accuracy.
Clinical Implications
The use of smartphone-based AI for ocular malignancy screening may facilitate earlier diagnosis and treatment.
Conclusion
The CaptureTumor platform represents an advancement in the early detection of ocular surface malignancies.
