PIUR tUS Infinity

The PIUR tUS Infinity system is a computer-aided detection tool designed to assist medical professionals by generating 3D volumetric reconstructions of the thyroid from 2D ultrasound cineloops using a combination of an IMU sensor and specialized software. Its methodology involves machine learning algorithms for thyroid lobe segmentation and nodule characterization support, providing quantitative data and ACR TI-RADS scoring assistance to enhance the diagnostic workflow for adult patients with thyroid conditions.

PIUR tUS Infinity is a computer-aided detection system designed to enhance the diagnostic workflow for thyroid and thyroid nodule assessment using ultrasound imagery. The system utilizes computer vision and machine learning algorithms to process sequences of 2D ultrasound images acquired from standard ultrasound machines. Key hardware components include the Infinity Box, which wirelessly transmits images from the ultrasound system, and the Infinity Sensor, an IMU-equipped device attached to the transducer to track its orientation. This combined information allows the Infinity Workstation software to generate tomographic 3D ultrasound volumes of the thyroid. The software then identifies the thyroid lobe, proposes its margins for user confirmation, and calculates its volume. Users can also mark, quantify, and visualize thyroid nodules in 3D, including multiplanar reconstructions. The system assists clinicians by providing tools for detailed analysis, including the calculation of the ACR TI-RADS level for nodules based on user input of five sonographic parameters. All system-proposed segmentations and classifications require user verification and potential adjustment before being incorporated into an automatically generated clinical report. Performance validation demonstrated the system's non-inferiority to conventional methods in both volume measurements (using metrics like Lin’s Concordance Correlation Coefficient and Mean Squared Error) and ACR TI-RADS classification (using Weighted Cohen’s Kappa coefficient across a diverse patient population and multiple ultrasound scanner models). Furthermore, the system was shown not to degrade ultrasound image quality and has undergone usability, electrical safety, and EMC testing. It is intended as an adjunctive tool for adult patients (22 years and older) and is not meant for direct patient contact or as a standalone diagnostic solution.

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. Given the performance data provided for volume measurement and ACR TI-RADS classification, what specific scenarios or nodule characteristics might still pose challenges to the system's accuracy where further investigation would be beneficial?; The document mentions machine learning algorithms for segmentation and proposal; what further details on the architecture or training datasets for these algorithms would clarify their robustness and generalizability beyond the reported patient cohort?; Considering the system's reliance on user verification of automated proposals, what studies could elucidate the impact of this human-in-the-loop approach on overall diagnostic time and consistency compared to fully manual or fully automated systems?; The document states non-inferiority across 7 ultrasound models; how sensitive is the 3D reconstruction and feature analysis to variations in image quality or settings from ultrasound systems not included in this validation set?; While the system calculates an ACR TI-RADS level from user input, how does the system guide or constrain the user's selection of the five ACR TI-RADS parameters to ensure standardized application and what impact does this have on the calculated score's reliability?