Us2.v2

Us2.v2 is an image post-processing software that utilizes machine learning algorithms for automated view classification and border detection in adult transthoracic cardiac ultrasound images. Its purpose is to provide quantitative measurements of cardiac structures and function, including dimensions, volumes, and systolic/diastolic parameters from B-mode, M-mode, and Doppler modalities, thereby aiming to enhance clinical echocardiographic workflow efficiency and support diagnostic review.

Us2.v2 is an advanced image post-processing analysis software designed for viewing and quantifying cardiovascular ultrasound images presented in DICOM format, specifically for adult patients. Its primary function is to automatically process acquired transthoracic cardiac ultrasound images, analyzing and making measurements to provide automated estimations of numerous cardiac structural and functional parameters. These parameters include left/right atrial and ventricular linear dimensions, volumes, systolic function, and diastolic function, measured using B-mode, M-mode, and Doppler (PW, CW, tissue) modalities. The software aims to support qualified healthcare practitioners in clinical decision-making by improving the echocardiographic workflow, enabling faster and more precise report generation while allowing clinicians full control over the final output, including manual markup options and report editing. The core technology of Us2.v2 relies on machine learning-based view classification and border detection to automate these echocardiographic analyses, covering the minimum standard dataset for adult echocardiograms. It is explicitly stated that the software does not report measurements associated with intracardiac lesions or complex adult congenital heart disease. Performance validation was conducted through bench studies comparing Us2.v2's automated analysis against manual analysis by trained echocardiography technicians or cardiologists, using Root Mean Square Error (RMSE) and Intraclass Correlation Coefficient (ICC) as statistical metrics. These tests utilized datasets from diverse US-based clinical cohorts and data from eight different ultrasound device vendors, ensuring segregation from algorithm training data, and included specific comparisons for Left Ventricular Strain analysis against the GE EchoPac system.

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 between various cardiac views and ensure accurate border detection in images with suboptimal quality or common anatomical variations?; What are the specific performance metrics (RMSE and ICC values) of Us2.v2 for each of the key measured cardiac parameters (e.g., LVEF, diastolic function parameters, atrial volumes) when compared directly against expert manual analysis across diverse patient demographics and various ultrasound systems?; How does the software handle and flag potential inaccuracies in measurements resulting from significant cardiac arrhythmias or the presence of pacemakers, which can impact image quality and measurement consistency?; What is the sensitivity and specificity of the integrated image quality assessment algorithms in identifying and filtering out images that would likely lead to clinically significant errors in automated measurements?; Beyond Left Ventricular Strain, what is the comparative performance and feature set of Us2.v2 against other commercially available automated echocardiographic analysis software for a comprehensive range of parameters?