Our commitment to responsible AI, physician autonomy, and patient safety in neonatal care.
We firmly believe in the "Human-in-the-Loop" approach. ROP Detect AI is designed to act as a second opinion and a triage tool for ophthalmologists. The final diagnosis and treatment decision must always remain with the qualified medical professional. Our system flags potential risks; it does not prescribe treatment.
We prioritize patient safety above all. The model has been calibrated to minimize False Negatives (missing a sick case) even if it increases False Positives, ensuring that no at-risk infant is overlooked during screening.
We are actively working to train our models on diverse datasets representing different populations and imaging devices to prevent algorithmic bias and ensure equitable healthcare outcomes for all neonates.
We avoid "Black Box" AI. Wherever possible, we utilize Class Activation Mapping (CAM) to highlight the regions of the retina that influenced the AI's decision, allowing doctors to understand the "Why" behind the "What".