Velacur

The Velacur system is designed to non-invasively quantify liver tissue stiffness and ultrasound attenuation using mechanically induced shear waves and volumetric ultrasound imaging. It incorporates deep learning algorithms for liver segmentation and shear wave quality assessment to enhance the accuracy and reliability of these measurements for clinical management of liver disease.

The Velacur system, model LI-1005, is a portable device designed for the non-invasive measurement of liver tissue stiffness and ultrasound attenuation to aid in the clinical management of patients with liver disease. Its methodology involves an activation unit, placed under a supine patient, which generates shear waves at frequencies of 40, 50, and 60 Hz within the liver. Concurrently, an ultrasound transducer, positioned over the patient's intercostal space, performs volumetric scans of the liver as these shear waves propagate. The system then calculates tissue stiffness by measuring shear wave speed (in the 40-70 Hz range) and determines the coefficient of attenuation by assessing the loss in power of the ultrasound beam. Key advancements in the Velacur system include the integration of two deep learning-based algorithms: one dedicated to organ/liver segmentation and another for evaluating shear wave quality, which replaces a prior rules-based system for scan quality. These algorithms are engineered to enhance the detection of high-quality shear waves and improve the identification of liver tissue, thereby facilitating more reliable data collection for the device's core elasticity and attenuation calculations. The AI components were trained using over 5,000 patient images per algorithm and subsequently validated on a separate dataset of over 1,500 images from 35-40 patients and volunteers, encompassing diverse demographics and various liver conditions. Performance metrics such as Dice Coefficient (≥0.7), pixel accuracy, sensitivity, and specificity (all ≥80%) were used, with ground truth established by expert manual segmentation and strict separation between training and testing datasets.

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 do the Velacur's deep learning-based organ segmentation and shear wave quality algorithms specifically improve the accuracy and reproducibility of stiffness and attenuation measurements compared to systems relying solely on rule-based or manual assessments?; What is the quantitative range of liver stiffness (kPa) and attenuation (dB/cm/MHz) that the Velacur system can accurately measure, and how does its performance vary across different etiologies and stages of liver disease, particularly in patients with high BMI or significant steatosis?; Given that the AI training data included patients primarily from hepatology clinics with more severe liver disease, what is the expected performance and robustness of the algorithms when applied to a general population or individuals with early-stage or milder liver conditions?; What were the specific acceptance criteria for the Dice coefficient, pixel accuracy, sensitivity, and specificity during the AI algorithm validation, and how were these thresholds determined to be clinically meaningful for liver assessment?; Can the Velacur system differentiate between inflammation and fibrosis when reporting tissue stiffness, or are there specific operational guidelines or complementary diagnostic data recommended to make this distinction in clinical practice?