Clarius Bladder AI

Clarius Bladder AI is a software application that utilizes artificial intelligence and non-adaptive machine learning algorithms to provide semi-automatic, non-invasive measurements of bladder volume from ultrasound images acquired by Clarius Ultrasound Scanners. Its scientific methodology relies on AI-powered image segmentation for the automated detection of bladder dimensions and subsequent volume calculation, with its accuracy and reliability validated as non-inferior to manual measurements by clinical experts through rigorous clinical studies.

Clarius Bladder AI is an ultrasound image processing software application that employs artificial intelligence, specifically non-adaptive machine learning algorithms, to facilitate semi-automatic, non-invasive measurements of bladder volume. Integrated within the Clarius App software, it is designed for use by trained healthcare professionals with Clarius Ultrasound Scanner systems (curvilinear and phased array transducers) on adult patients. The system automatically detects and measures bladder depth, width, and height from sagittal and transverse ultrasound views, calculates the volume, and places calipers. Users retain the option to manually adjust these AI-generated measurements, ensuring that clinical judgment guides final assessments. The software is compatible with iOS and Android operating systems. The device's safety and effectiveness were substantiated through extensive non-clinical and clinical testing. Non-clinical evaluations confirmed compliance with standards like IEC 62304 (software lifecycle) and ISO 14971 (risk management), including cybersecurity measures. Clinical performance was verified through retrospective and prospective studies comparing AI-driven measurements to manual measurements by expert clinicians. Both studies demonstrated that Clarius Bladder AI's automated measurements are non-inferior to expert manual measurements, showing strong agreement (evaluated using Intraclass Correlation Coefficients, dice scores, and Jaccard index) with mean differences within a 25% equivalence margin. A subsequent clinical validation study confirmed the device meets user needs and performs as intended in representative clinical environments for semi-automated bladder volume assessment.

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. What specific deep neural network architecture is utilized for the image segmentation and view classification, and how was it optimized during the tuning phase using the validation dataset?; How does the Clarius Bladder AI algorithm's performance vary across a wider spectrum of bladder volumes, particularly at the extreme ends (e.g., very low or very high volumes), beyond the mean differences reported?; Considering the gender distribution in the training dataset (999 male, 353 female) and unknown ethnicities in the retrospective study, what specific measures were taken to validate the model's performance and ensure equitable accuracy across diverse demographic subgroups not proportionally represented?; The device automatically detects sagittal versus transverse views; what is the documented accuracy of this view classification component, and what occurs if an image view is ambiguous or misclassified by the AI?; What were the precise criteria for defining 'qualified experts' or 'clinical truthers' in the retrospective and prospective studies, and what was the measured inter-observer variability among these experts for manual bladder volume measurements, which served as the ground truth?