HM70 EVO Diagnostic Ultrasound System

The HM70 EVO Diagnostic Ultrasound System is a versatile, mobile imaging platform utilizing multiple ultrasound modalities, including Doppler, 3D/4D, and elastography, to acquire diagnostic images and analyze body fluids for a wide array of clinical applications such as obstetrics, cardiology, and dermatology. Its scientific basis involves advanced ultrasound physics coupled with software-controlled image processing and AI-driven tools like UterineAssist and NerveTrack, which have demonstrated specific quantitative performance in anatomical segmentation, measurement accuracy, and nerve tracking to enhance diagnostic utility.

The HM70 EVO Diagnostic Ultrasound System is a mobile, software-controlled platform designed for general-purpose diagnostic ultrasound imaging and body fluid analysis across numerous clinical specialties, including fetal/obstetrics, abdominal, cardiac, musculoskeletal, and the newly added dermatology application. It functions by acquiring ultrasound data via various probes, including the new L3-22 transducer and the migrated LA3-22AI transducer, and displays this information through a comprehensive suite of imaging modes such as 2D, M-mode, Color Doppler, Power Doppler, Pulsed and Continuous Wave Doppler, Tissue Doppler, 3D/4D imaging, and ElastoScan. The system also provides tools for anatomical measurements and analysis, and features real-time display of mechanical and thermal acoustic output indices, along with accessories like a new gel warmer and an additional portable power supply. Key technological advancements in the HM70 EVO include the integration of AI-powered software functions: NerveTrack and UterineAssist. UterineAssist, validated using 450 sagittal and 150 transverse uterine images from Korean females, demonstrated high performance in image segmentation (average dice-scores of 96% for uterus and 92% for endometrium) and accurate feature point extraction (errors ≤ 5.8mm for uterus, ≤ 4.3mm for endometrium) and size measurements (errors ≤ 2.0mm). The expanded NerveTrack feature was validated for detecting 10 different nerve types using 2,146 images from 18 Korean subjects, achieving an average accuracy of 91.50% and a speed of 3.71 FPS, with good generalizability across BMI subgroups. These non-clinical tests indicate the system's enhanced capabilities for diverse diagnostic procedures.

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 UterineAssist, particularly its segmentation accuracy (dice-scores of 96% for uterus, 92% for endometrium) and measurement error (<2.0mm), compare when applied to uterine images from a more ethnically diverse population or patients with significant uterine pathologies not extensively covered in the Korean dataset?; What is the reported impact of the NerveTrack AI feature (91.50% accuracy, 3.71 FPS speed) on procedure time and diagnostic confidence in clinical settings, particularly for users with varying levels of experience in ultrasound-guided nerve procedures across the 10 validated nerve types?; Can you provide specific image quality comparisons or quantitative data (e.g., resolution, penetration depth) demonstrating the advantages of the new L3-22 and migrated LA3-22AI transducers, especially for the newly added dermatology application and for imaging challenging patient anatomies?; Given that UterineAssist was tested on data from four specific Samsung Medison ultrasound systems, how robust is its performance when images are acquired from different ultrasound machines or under variable acquisition protocols?; What are the computational requirements for the AI features (NerveTrack, UterineAssist), and how do they affect the overall system responsiveness and workflow efficiency, especially when multiple advanced modes are used concurrently?