Bladder AI (AIBV01)

Bladder AI is a software medical device designed to aid qualified users in quantifying bladder volume from ultrasound images for patients aged two years and older. It employs machine learning algorithms for semi-automatic landmark placement and image segmentation to derive bladder dimensions and volume from DICOM ultrasound images.

Bladder AI is a standalone software as a medical device (SaMD) developed by Exo Inc. to assist qualified healthcare providers with the image-based assessment of bladder ultrasound images in patients aged two years or older. The software is designed to simplify clinical workflow by helping users evaluate, quantify bladder volume, and generate reports. It processes imported Digital Imaging and Communications in Medicine (DICOM) images from ultrasound scanners, allowing for the measurement of bladder volumes from both single-frame and multi-frame ultrasound images. Key functionalities include tools for viewing ultrasound images of the bladder, machine learning-based semi-automatic landmark placements for bladder dimension and volume measurements, and the creation of examination reports. The core technology utilizes machine learning for image segmentation to detect bladder borders, enabling semi-automatic landmark placement which can be modified by the user. The clinical performance was validated using data from 122 subjects, encompassing diverse demographics (ages 2-95 years) and images acquired from various cart-based and portable ultrasound devices (frequency 1.3-9 MHz) with bladder volumes ranging from 11 to 645 mL. Performance was benchmarked against the average bladder volume measurements from three expert clinicians. The results demonstrated high accuracy and reliability, with an Intraclass Correlation Coefficient (ICC) of 0.98 for dual-view measurements (mean volume difference 2 mL, Limits of Agreement -42 to 46 mL) and an ICC of 0.97 for single-view measurements (mean volume difference 3 mL, Limits of Agreement -49 to 55 mL). The algorithm's performance was consistent across subgroups based on age, gender, BMI, and ultrasound device manufacturers, supporting its generalizability.

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 machine learning algorithm differentiate bladder tissue from surrounding structures, particularly in images with artifacts or patients with anatomical variations?; What was the inter-operator variability when using Bladder AI compared to manual measurements by expert clinicians in the validation study?; How does the software's performance vary with different ultrasound machine manufacturers and image acquisition settings (e.g., frequency, depth)?; What is the impact of Body Mass Index (BMI) on the accuracy of Bladder AI, and was this specifically stratified in the performance data?; Are there plans to expand the algorithm's capabilities, such as detecting other bladder abnormalities or integrating with specific EMR/PACS systems for seamless reporting?