EPIQ Series Diagnostic Ultrasound Systems; Affiniti Series Diagnostic Ultrasound Systems

The Philips EPIQ and Affiniti ultrasound systems are enhanced with new software, Smart View Select (SVS) and Segmental Wall Motion (SWM), aiming to streamline and improve cardiac assessments from transthoracic echocardiograms. SVS employs AI for intelligent cardiac view selection for left ventricular analysis, while SWM utilizes machine learning for automated quantification of regional wall motion and calculation of an overall Wall Motion Score Index, with both functionalities validated by strong correlations to reference methods in nonclinical evaluations.

This document details software enhancements for the Philips EPIQ and Affiniti Series Diagnostic Ultrasound Systems, introducing Smart View Select (SVS) for the EPIQ series and Segmental Wall Motion (SWM) for both. SVS automates the selection of optimal cardiac images from Adult Transthoracic Echocardiography (TTE) for left ventricle (LV) analysis, utilizing a Deep Learning AI for view classification and a non-AI algorithm that considers view and image depth (preferring shallower scans >10cm) for selecting appropriate A4C, A2C, and A3C views. The SWM application automatically evaluates regional LV function from adult TTE exams, performing border detection and tracking on three apical views, and then uses machine learning to assign wall motion scores to 17 LV segments and calculate an overall Wall Motion Score Index (WMSI). These software features are integrated into existing systems without hardware changes, using established TTE transducers, and build upon algorithms from reference devices (LVivo Seamless for SVS, LVivo SG for SWM) with minor modifications, such as a backup border initialization for SWM and user-editable segment scores. Nonclinical performance was evaluated through retrospective studies. The SVS feature demonstrated strong correlation between automated and manual expert clip selection for Global Longitudinal Strain (GLS Pearson's r=0.911) and Ejection Fraction (EF Pearson's r=0.941) outputs. The SWM feature showed high correlation for WMSI compared to a reference method, both with manually selected clips (Pearson's r=0.957) and SVS-selected clips (Pearson's r=0.913).

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 AI in Smart View Select handle variations in image quality or atypical cardiac anatomy during view classification?; What is the clinical impact of the heuristic logic in SVS that prioritizes shallower scan depths (above 10cm) on the accuracy of subsequent LV analysis, particularly in patients with larger body habitus?; For the Segmental Wall Motion software, what are the specific performance differences, if any, when using the primary border initialization algorithm (from K161382) versus the backup AutoStrain LV algorithm?; How does the ability for users to manually edit SWM scores via a drop-down menu, rather than adjusting borders, affect inter-observer variability and the potential for bias in clinical reporting?; Given the strong correlation results in the retrospective studies, are there plans for prospective studies to evaluate the real-world clinical utility and workflow efficiency improvements provided by SVS and SWM in diverse patient populations?