HS40 Diagnostic Ultrasound System

The HS40 Diagnostic Ultrasound System is a versatile imaging platform intended for a wide range of clinical diagnostic applications, employing multiple ultrasound modalities for image acquisition and fluid analysis. Its methodology is enhanced by artificial intelligence through features like UterineAssist for automated gynecological assessment and an expanded NerveTrack for precise nerve identification, both of which have demonstrated strong performance metrics in non-clinical validation tests.

The HS40 is a general-purpose, mobile, software-controlled diagnostic ultrasound system designed to acquire ultrasound data and facilitate the analysis of body fluids. It functions by displaying this data through a comprehensive suite of imaging modes, including 2D, M-mode, Color Doppler, Power Doppler, Pulsed and Continuous Wave Doppler, Harmonic imaging, Tissue Doppler Imaging and Wave, Panoramic Imaging, 3D/4D imaging, and ElastoScan. These capabilities enable operators to measure anatomical structures and utilize analysis packages to support clinical diagnosis across a wide spectrum of applications such as fetal/obstetrics, abdominal, gynecological, pediatric, small organ, cephalic, trans-rectal, trans-vaginal, musculoskeletal, urological, cardiac, and peripheral vessel imaging. The system also features real-time display of mechanical and thermal acoustic output indices for operational awareness. This iteration of the HS40 system incorporates significant technological advancements, notably the introduction of UterineAssist and an expanded NerveTrack, both AI-based features. UterineAssist, leveraging deep learning, automates uterine and endometrial segmentation with high accuracy (average dice-scores of 96% for uterus and 92% for endometrium), extracts feature points with errors of 5.8mm or less for the uterus and 4.3mm or less for the endometrium, and performs size measurements with errors of 2.0mm or less, based on tests primarily on Korean subjects. The NerveTrack feature has been enhanced to detect a wider range of nerve types in the elbow (MN, UN, RN) and neck/shoulder (ISBP, SCBP, AxBP), demonstrating an average accuracy of 91.7% and a processing speed of 7.93 FPS in validation studies also conducted on a Korean dataset with varying BMIs. Additional updates include the integration of new transducers (LA2-9S, PA3-9B, PA1-5A, CA1-7AD), a modification to the EV2-10A transducer's housing material, and revised caster specifications.

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 performance of the UterineAssist feature vary with different uterine pathologies (e.g., fibroids, adenomyosis, polyps) or significantly altered uterine anatomy?; What were the specific performance metrics (accuracy, sensitivity, specificity) for each of the newly added nerve types in the expanded NerveTrack feature, particularly for nerves at varying depths or in patients with high BMI?; Considering the validation data for UterineAssist and NerveTrack was based on an all-Korean population, what are the plans for validating these AI features across more diverse ethnic groups and body habitus types to ensure broader generalizability?; What is the quantitative impact of the new transducers (LA2-9S, PA3-9B, PA1-5A, CA1-7AD) on image resolution, penetration depth, and overall diagnostic confidence for their intended applications compared to previously available transducers for the HS40 system?; Beyond anatomical measurement, does the UterineAssist feature provide any qualitative assessment or classification of endometrial patterns or uterine textural changes that could aid in diagnosing specific conditions?