Benchtop Calibration of an Image-Based Algorithm for Objective Quantification of Menstrual Blood Loss

Background Accurate quantification of menstrual blood loss (MBL) is essential for diagnosing and managing heavy menstrual bleeding, however, current clinical practice relies largely on subjective self-reporting. Image-based algorithms offer a promising objective alternative but often require clinical validation for each specific menstrual product. This study aimed to evaluate whether a benchtop model using simulated menstrual blood could reliably calibrate a clinically validated image-based MBL quantification algorithm, enabling product validation without additional clinical trials. Methods Porcine whole blood was diluted with 0.9% saline to create eight simulated menstrual blood solutions spanning physiologically relevant hemoglobin concentrations and blood fractions. Solutions were applied in 5 mL increments to two commercially available menstrual pad types up to their maximum absorbency. Pads were imaged using a standardized mobile imaging protocol and processed through the Quantiflow image-analysis pipeline. Predicted blood volumes were compared with known applied volumes using linear regression, error analysis, and Bland–Altman agreement testing to identify the dilution condition that best matched the algorithm’s clinical training data. Results Across all conditions, predicted and true blood volumes demonstrated strong linear relationships (R² = 0.907–0.976). Undiluted porcine whole blood showed the closest alignment with the algorithm’s clinical performance, yielding a slope of 0.904 and R² of 0.976 for volumes ≤ 20 mL. Prediction error decreased exponentially with increasing hemoglobin concentration and blood fraction. Bland–Altman analysis for the optimal condition demonstrated a mean bias of − 0.55 mL, with 90% limits of agreement from − 4.32 to 3.22 mL. Error metrics and repeatability analyses confirmed high accuracy and reproducibility within physiologic ranges. Conclusion A benchtop model using undiluted porcine whole blood provides a reproducible and accurate method for calibrating image-based menstrual blood loss algorithms. This approach enables efficient validation across menstrual products while reducing reliance on human clinical studies, supporting scalable and ethical development of objective menstrual health assessment tools.