AVA (Augmented Vascular Analysis)

AVA (Augmented Vascular Analysis) is a software leveraging artificial intelligence and machine learning to analyze DICOM-compliant vascular ultrasound images of carotid and lower limb arteries. Its methodology encompasses automated segmentation for intima-media thickness measurement, Doppler velocity extraction, text recognition, and waveform classification, aiming to produce comprehensive vascular reports for clinician review and approval.

See-Mode AVA (Augmented Vascular Analysis) is a standalone software designed for the analysis, measurement, and reporting of DICOM-compliant vascular ultrasound images, specifically focusing on carotid and lower limb arteries. It operates on standard computer hardware and is accessed via a web browser, utilizing proprietary algorithms incorporating artificial intelligence and machine learning for its analytical capabilities. The software processes B-Mode ultrasound images to segment vessel walls and measure the intima-media thickness (IMT) of the carotid artery. For Doppler images, it identifies peak systolic and end diastolic velocities (PSV and EDV) and reads image annotations using text recognition. Furthermore, it analyzes lower limb artery Doppler scans to detect waveform types (monophasic, biphasic, or triphasic) and generate qualitative representations of stenosis based on velocity ratios. The core science behind AVA involves advanced image processing techniques, including neural networks for IMT segmentation and signal processing for Doppler waveform analysis. Performance data from retrospective studies demonstrated high efficacy: IMT measurements showed a strong correlation (0.89) with expert readers, outperforming a predicate device. The text recognition algorithm achieved 92-96% accuracy in reading image annotations. Analysis of Doppler waveforms for PSV and EDV yielded correlation coefficients of 0.98 and 0.97, respectively, against clinician-annotated values. For lower limb Doppler images, the waveform type classifier demonstrated 93% accuracy compared to expert assessments. The software generates a vascular ultrasound report from these analyses, which requires review and approval by a qualified clinician before being used adjunctly for patient cardiovascular health assessment.

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 AI-driven image segmentation for IMT measurement in AVA compare in terms of processing time and inter-observer variability reduction versus traditional manual or semi-automated methods?; What specific machine learning models are employed for vessel wall segmentation and IMT measurement, and how were these models trained and validated to ensure generalizability across diverse patient populations and ultrasound machine variability?; The performance data indicates text recognition accuracy of 92-96% and waveform classification accuracy of 93%; what are the clinical implications of the remaining error rates, and are there specific scenarios where the software is more prone to misinterpretation?; For lower limb artery analysis, how does the software's qualitative drawing of stenosis based on Doppler velocity ratios correlate with quantitative angiography or other imaging modalities in assessing stenosis severity?; Given its cloud-based platform, what specific measures are in place to ensure data security, patient privacy (e.g., HIPAA compliance), and system uptime/reliability, particularly in critical diagnostic workflows?