Caption Guidance

Caption Guidance is an AI-driven software accessory for ultrasound systems, intended to assist medical professionals in acquiring standardized, diagnostic-quality 2D transthoracic echocardiograms in adults by providing real-time, expert-emulating guidance. Its methodology relies on pre-trained algorithms for image assessment and user instruction, with future software updates managed through a predetermined change control plan ensuring algorithms are locked prior to deployment.

Caption Guidance is a software-only device designed to assist medical professionals in acquiring cardiac ultrasound images. It functions as an accessory to compatible general-purpose diagnostic ultrasound systems and aims to help users obtain anatomically correct images representing standard two-dimensional (2D) transthoracic echocardiographic (2D-TTE) diagnostic views for adult patients. The software is indicated for use in acquiring ten specific standard views, including Parasternal Long-Axis, Apical 4-Chamber, and Subcostal views. The system employs artificial intelligence to emulate the expertise of sonographers, providing real-time guidance through features like a 'Quality Meter' for image quality feedback, 'Prescriptive Guidance' for transducer manipulation instructions, 'Auto-Capture' for automatic recording of diagnostic quality clips, and 'Save Best Clip' for retrospective saving of the best image obtained. The technological basis is identical to a predicate device, with this iteration focusing on modifications under a Predetermined Change Control Plan (PCCP). This plan outlines how anticipated software modifications, specifically to algorithms, will be trained, tuned, and locked before field release, explicitly stating that the PCCP does not include provisions for adaptive algorithms that learn continuously in the field.

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 'Quality Meter' quantitatively define and measure the diagnostic quality of an ultrasound clip in real-time across diverse patient anatomies?; What specific AI models or algorithms are utilized in the 'Prescriptive Guidance' feature to emulate sonographer expertise, and what was the diversity of the dataset used for their training?; What are the performance metrics (e.g., acquisition time, required user expertise, image quality scores) of the Caption Guidance software when used by medical professionals with varying levels of ultrasound experience compared to expert sonographers?; The Predetermined Change Control Plan (PCCP) outlines methods for future modifications; what are the specific validation and verification processes to ensure these algorithmic updates maintain or improve performance and safety without introducing bias, especially across different compatible ultrasound systems?; How does the 'Auto-Capture' feature differentiate between transient high-quality frames and sustained diagnostic-quality views, and what is its sensitivity and specificity for capturing clinically relevant images?