Clarius OB AI

Clarius OB AI is an AI-powered software designed to assist healthcare professionals by automatically segmenting fetal anatomy and measuring key biometric parameters from ultrasound images acquired with the Clarius Ultrasound Scanner. Its methodology relies on a deep learning algorithm validated through non-inferiority studies against expert clinicians, aiming to enhance efficiency and consistency in obstetric assessments, particularly for estimating gestational age, while allowing for user oversight and manual adjustments.

Clarius OB AI is an assistive software tool integrated into the Clarius Ultrasound App, designed for use with the Clarius Ultrasound Scanner system by trained healthcare professionals. Its purpose is to aid in obstetric ultrasound imaging by performing non-invasive, semi-automated measurements of key fetal biometric parameters on adult patients. The methodology involves a deep learning image segmentation algorithm that, upon selection of an obstetric preset (OB or Early OB), automatically detects and segments fetal anatomy (head, abdomen, femur, or whole fetus for early OB) and places calipers to measure head circumference (HC), abdominal circumference (AC), femur length (FL), bi-parietal diameter (BPD), and crown rump length (CRL). The science relies on a machine learning model trained on diverse ultrasound data to provide these measurements in B-Mode, with users retaining the ability to manually adjust the AI-generated calipers and the ultimate responsibility for clinical interpretation. The performance of Clarius OB AI was validated through non-clinical testing against relevant industry standards and extensive clinical performance testing. A retrospective clinical verification study involving 347 subjects and images from diverse international sites demonstrated that Clarius OB AI's automated measurements were non-inferior to manual measurements performed by expert clinicians, with statistically significant p-values (<2.2e-16) and high intraclass correlation (ICC of 0.99). Segmentation accuracy was confirmed with high Jaccard (range 0.73-0.94) and Dice (range 0.84-0.97) scores. A subsequent clinical validation study confirmed its usability and that it performs as intended in a representative user environment, meeting user needs for semi-automated fetal biometric measurements used in calculating fetal gestational age. The device also includes a Predetermined Change Control Plan for future AI model updates and improvements, addressing aspects like model architecture, training methods, post-processing algorithms, and compatibility with new data input sources.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the Clarius OB AI deep learning model vary across different gestational age ranges, particularly at the extremes (e.g., very early OB vs. late third trimester)?; What specific image quality parameters are most critical for the Clarius OB AI to achieve optimal segmentation and measurement accuracy, and how does the system alert users to suboptimal image quality?; Given the diverse ethnic groups in the training and verification datasets, are there any performance differences or biases observed for specific fetal biometric parameters in distinct populations, and how are these addressed?; The Predetermined Change Control Plan mentions mitigating overfitting and unintended bias; what specific techniques beyond regularization and cross-validation are employed, particularly for ensuring fairness across diverse patient demographics with new data input sources?; How does the semi-automated workflow with Clarius OB AI, including the option for manual adjustment, impact overall examination time and inter-observer variability compared to fully manual measurement techniques in real-world clinical settings?