Auto 3D Bladder Volume Tool

The Auto 3D Bladder Volume Tool is a software application for Butterfly Network ultrasound systems that provides non-invasive bladder volume measurements in adults to support physician diagnosis. It utilizes a U-Net artificial intelligence algorithm to automatically segment 2D B-mode ultrasound images and calculate bladder volume through integration.

The Auto 3D Bladder Volume Tool is a software application designed for use with Butterfly iQ Ultrasound Systems to perform non-invasive measurements of bladder volume in adult populations, aiding physicians in diagnosis. Operating in B-mode with a specific "Bladder preset," the tool guides clinicians to acquire 25 two-dimensional (2D) ultrasound images of the bladder. Once these images are captured, the software automatically calculates an estimated bladder volume, which is then presented to the clinician. This quantified data serves as supplementary information within standard diagnostic pathways and is intended for use by qualified medical professionals. The tool is specifically contraindicated for fetal or pediatric use, pregnant patients, patients with ascites, or those with open skin or wounds in the suprapubic region. The technological foundation of the Auto 3D Bladder Volume Tool is an artificial intelligence-driven approach, specifically employing the U-Net convolutional network for automatic image segmentation. The bladder volume is subsequently computed through a process of integration over these segmented binary images from the 25 acquired 2D scans. Butterfly Network conducted comprehensive software verification and validation, alongside performance testing that included a non-interventional validation study where the primary endpoint was met. This study used a patient dataset representative of typical clinical scenarios regarding volume range, gender, and age. Inter and intra-operator variability in processing images was assessed, and variability testing also demonstrated consistent performance across different image clips and frames from the same patient, establishing the tool meets its design specifications and intended use. The tool has a measurement accuracy of ±7.5mL for volumes 0-100mL and ±7.5% for volumes 100-740 mL, within a measurement range of 0 to 740 mL.

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 U-Net algorithm's performance in segmenting the bladder compare across different patient demographics (e.g., varying BMIs, bladder morphologies, or presence of comorbidities affecting bladder shape) not explicitly detailed in the non-interventional study?; What is the impact of operator variability in acquiring the initial 25 2D images on the accuracy and consistency of the automated volume calculation, beyond the inter/intra-operator variability assessed for image processing?; The tool acquires 25 2D images; how was this number optimized for accuracy versus acquisition time, and what is the sensitivity of the volume measurement to a lower or higher number of images?; Given the contraindication for patients with ascites, what is the specific mechanism by which ascites interferes with the U-Net algorithm's accuracy, and are there potential software adjustments or image processing techniques that could mitigate this?; What are the measurement accuracy differences, if any, when the Auto 3D Bladder Volume Tool is used with different models or generations of Butterfly iQ Ultrasound Systems, particularly concerning transducer variations or image processing capabilities inherent to the hardware?