LVivo Software Application

The LVivo Software Application provides automated, non-invasive analysis of ultrasound DICOM images to quantify cardiac (left and right ventricular) and bladder parameters. It employs image processing, Deep Learning Neural Networks, and machine learning with active contour methods to derive measurements such as ejection fraction, strain, right ventricular size and function, and bladder volume.

The LVivo Software Application is a platform intended for the non-invasive processing of ultrasound images to detect, measure, and calculate relevant medical parameters of anatomical structures and their function. It analyzes echocardiographic DICOM movies, using two orthogonal planes (4-chamber and 2-chamber views) for automated analysis of Left Ventricular (LV) function, including global ejection fraction (EF) and strain. The platform also features modules for Right Ventricle (LVivoRV) evaluation and bladder volume measurement (LVivo Bladder), operating on Windows/Linux systems with an Android option for LVivo EF. The LVivoRV module utilizes image processing and Deep Learning Neural Networks to automatically assess RV function from apical 4-chamber views, yielding measurements such as End-Systolic Area (ESA), End-Diastolic Area (EDA), Fractional Area Change (FAC), TAPSE, S', and Free Wall Strain. Clinical validation demonstrated good correlation between LVivoRV and manual measurements for FAC (r=0.79), EDA (r=0.92), and ESA (r=0.93). The LVivo Bladder module employs machine learning and active contour algorithms to automatically segment bladder contours from sagittal and transverse ultrasound views for volume calculation. Clinical studies showed excellent agreement (Kappa=0.84) with manual methods for differentiating post-voiding residual volumes around a 200mL threshold and a high correlation (r=0.94) for overall volume calculation compared to manual tracing.

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 Deep Learning Neural Network in LVivoRV handle variability in ultrasound image quality or atypical patient anatomy, and what are its performance characteristics in such scenarios?; The LVivo Bladder module allows sonographer adjustments to calipers; what is the typical impact of these manual adjustments on the final bladder volume accuracy compared to the fully automated output?; For LVivoRV's TAPSE and Free Wall Strain measurements, what were the specific sonographic or clinical characteristics of the 'outlier cases' whose omission significantly improved correlations, and how does the system address such cases in routine use?; How does the LVivoRV module's automated assessment of right ventricular function, particularly parameters like Fractional Area Change and strain, compare in accuracy and reproducibility to cardiac MRI, often considered a reference standard for RV assessment?; Beyond the quantitative outputs, does the LVivo platform incorporate any mechanisms to provide users with a confidence score for its automated measurements or flag images of suboptimal quality that might compromise analytical accuracy?