Real-time quality control in optoacoustic mesoscopy for enhancing data quality and standardization in clinical studies

Raster scan optoacoustic mesoscopy (RSOM) has matured as a medical imaging modality that enables unique high-resolution visualization of optical contrast at depths of several millimeters. Compared with other optical methods, optoacoustics is less affected by photon scattering, enabling superior imaging of dermatological, cardiometabolic, and other conditions. A critical requirement for clinical adoption is the development of methodology that ensures quality control and standardization across subjects, time points, and acquisition environments. We present a machine-learning-based automated real-time quality control method for RSOM using signal-derived metrics for noise and motion. The model was trained and evaluated on 1725 clinical RSOM scans benchmarked against visually perceived image quality ratings from eight experts. The method enables real-time feedback during acquisition to identify suboptimal scans and support standardized high-quality data acquisition. We discuss the impact of the method on clinical RSOM deployment, and the cost benefits achieved through data standardization.