Female-RHINO AI automates uterine MRI analysis in under 70 seconds

A team of researchers from multiple institutions (led by Deepak Bhatia) has released Female-RHINO: a real-time, scanner-integrated AI framework for automated quantitative uterine MRI analysis. The system connects directly to the MRI scanner during image acquisition, running deep learning models that segment the uterus and surrounding structures, detect common incidental findings like fibroids and Nabothian cysts, and extract six anatomical landmarks for biometric assessment. All results are compiled into a structured clinician-oriented report with visualizations—no human interaction required.

Trained on over 500 multi-center datasets spanning diverse protocols, vendors, and patient populations, the models achieved strong performance: mean Dice similarity coefficients of 0.82 for the uterus and 0.80 for fibroids, with landmark detection accuracy within 3.7 mm radial error. End-to-end processing completes in under 70 seconds, meaning results are available while the patient is still in the scanner. Prospective deployment confirmed standardized, reproducible analyses with high inter-observer agreement.

Female-RHINO addresses long-standing challenges in uterine MRI—anatomical variability, observer dependence, and lack of automated tools. By embedding AI directly into the scanning workflow, it eliminates manual segmentation and reporting delays, enabling radiologists to focus on interpretation rather than measurement. This framework could set a new standard for automated pelvic imaging analysis in clinical practice.