Samsung Medison has launched the HERA Z10, a premium obstetric and gynecologic ultrasound system that brings several AI-assisted workflow and imaging features from the company's flagship HERA Z20 into a second model in the HERA Z Series.
The company presented the system at the International Society of Ultrasound in Obstetrics and Gynecology World Congress in London, held from September 4 to 7.
The most consequential part of the announcement is not that an ultrasound machine contains “AI.” It is how the software is being applied: automated acquisition and measurement steps, structure segmentation, and assistance with repetitive parts of an examination where workflow consistency matters.
Live ViewAssist targets the repetitive part of a scan
One of the HERA Z10's central software features is Live ViewAssist, which is designed to recognize standard ultrasound views and automate parts of the measurement workflow during an examination.
Samsung cites joint clinical research with Professors Giuseppe Rizzo and Daniele Di Mascio of Sapienza University of Rome and says Live ViewAssist reduced scan time during detailed mid-trimester ultrasound examinations by up to 52%.
That number should be read as a company-reported result from the referenced research context, not as a guarantee that every examination will be 52% shorter. Real-world performance can vary with patient characteristics, operator technique, clinical workflow and the exact protocol being used.
Still, the use case is plausible and important. Obstetric ultrasound requires clinicians to acquire and measure a series of standard anatomical views. Software that reliably identifies those views and reduces repeated manual steps can potentially shift more of the operator's time toward interpretation and difficult cases.
EzStructure automates another part of the measurement workflow
Samsung also highlights EzStructure, which automates manipulation involved in selected measurement parameters. The company says the feature can reduce manual manipulation by up to 71%.
Again, this is best understood as a workflow claim rather than a claim that diagnostic decision-making itself is 71% automated. The distinction matters: automating a measurement step is different from replacing the clinician who determines whether the image is adequate, interprets the findings and integrates them with the clinical situation.
That division of labour is likely to be a recurring pattern in medical AI. Some of the most useful systems may be those that quietly remove repetitive interface work while leaving consequential interpretation with trained professionals.
Imaging features extend beyond automated measurements
The HERA Z10 also includes MV-Flow 3D, which Samsung says is intended to visualise microvascular blood flow at high resolution. The company positions it for applications including assessment of placental and uterine or umbilical-cord vasculature in complex pregnancies.
Another feature, FibroidVue, uses AI to segment, visualise and measure the uterus, endometrium and uterine fibroids.
Segmentation is a familiar medical-imaging task for machine learning: identify the boundaries of a structure in an image so that measurements and visualisation can be generated more consistently. Its value depends heavily on the quality of the underlying image, the populations represented in validation data and the clinician's ability to recognise when an automated boundary is wrong.
The system still needs clinical evidence, not just feature names
Medical imaging is a domain where impressive automation demos are not enough. The questions that matter are clinical and operational: How often does the software make a useful suggestion? How often does it need correction? Does it perform similarly across patient groups, gestational ages and difficult acquisition conditions? Does saved scan time translate into better throughput or simply move work elsewhere?
Samsung's announcement describes the capabilities and reports selected workflow improvements, but it does not by itself establish broad clinical superiority over other ultrasound platforms.
For buyers and clinicians, the useful evidence will include peer-reviewed validation, regulatory clearances for specific functions and indications, multicentre performance, failure-mode information and experience from routine practice.
Why this launch matters
The HERA Z10 illustrates where AI is becoming embedded in medical devices: not as a single autonomous diagnostic model, but as a collection of narrowly scoped tools distributed through the imaging workflow.
That model can be valuable because ultrasound is highly operator-dependent. Assistance with view recognition, measurements and structure segmentation may improve efficiency and consistency without requiring the system to make the final clinical judgment.
The harder question is whether those gains remain reliable outside controlled studies and demonstrations. That is what will determine whether features such as Live ViewAssist and EzStructure become routine workflow infrastructure or remain premium-system differentiators.