Samsung AI Ultrasound R20: Experiences
The integration of Artificial Intelligence (AI) into medical diagnostics is progressing. Current developments such as the Samsung R20 ultrasound system and NVIDIA-based digital lung twins show how AI can support the clinical workflow and improve precision.
Source: Own Presentation
AI in Healthcare
In interviews and experience reports, radiologists and sonographers express a mixture of relief and skepticism regarding the use of AI. The expectation is that AI will simplify daily work without devaluing their own expertise. Current announcements, such as the Samsung R20-Ultraschall auf der RSNA 2025 and the NVIDIA-basierten Digital Twins der Lunge von L&T Technology Services, , show a shift away from isolated “gadget algorithms” towards AI assistance deeply integrated into the clinical workflow.
Radiology worldwide is under double pressure: an aging population and the increase in chronic diseases are raising the demand for imaging, while the number of radiologists is not growing proportionally. A recent analysis for the USA describes approximately 34.000 praktizierende Radiolog:innen, , of whom around a third are over 55 years old. This leads to a gap between open positions and incoming specialists. Similar bottlenecks are described in the British NHS, where cancer patients face extended waiting times for imaging diagnostics ( The Guardian).
Radiology is the field where AI software is most widespread. More than three-quarters of AI products approved by the US Food and Drug Administration in the medical sector are designed for radiological applications. Approximately two-thirds of US radiology departments already use some form of AI support ( The Washington Post). ). So far, detection algorithms for cerebral hemorrhages, lung nodules, or breast cancer dominate. However, tools that suggest protocols, sort worklists, or generate draft reports are increasingly being added ( The Washington Post, Nature).
Samsung R20 Ultrasound
Samsung Medison is introducing the new R20 ultrasound system at RSNA 2025 in Chicago. This system combines over a dozen AI functions for real-time support, automatic measurements, report assistance, and workflow automation ( Samsung Global Newsroom, TechBuzz.ai).
The core of the system is the “Advanced Imaging Engine”, which couples hardware and software-based beamforming. This enables higher image quality and diagnostic precision in difficult situations, for example, with obese or severely ill patients ( healthtechhotspot.com).
The R20's AI tools cover multiple levels: during the scan, examiners receive guidance on probe position, cut direction, and image quality. Recurring measurements such as distances, areas, or volumes are automatically detected and translated into structured report data ( healthtechhotspot.com, TechBuzz.ai). ). Thus, the R20 aims for practical support: fewer clicks, less manual remeasuring, and more consistency between examiners with different experience levels ( healthtechhotspot.com).
Source: news.samsung.com
The Samsung R20 ultrasound device in action.
Another important aspect is ergonomics. According to Samsung, up to 90% of ultrasound users report suffering from pain during scanning, caused by high probe pressure, abducted shoulders, or twisted posture ( healthtechhotspot.com). ). The R20 has been tested by an independent institute and meets 100% of recognized ergonomic guidelines. In combination with AI-supported workflows, this is intended to reduce physical strain and counteract the shortage of skilled workers ( healthtechhotspot.com, TechBuzz.ai).
Initial experiences with the R20 come mainly from hands-on sessions and press releases. It is emphasized there that the more than twelve AI tools are positioned as “practical clinical support” and not as a gimmick. This differs from previous generations of AI features, which often ran as an add-on alongside the actual workflow ( healthtechhotspot.com, TechBuzz.ai). ). Reliable independent studies or long-term experiences from large hospitals are still scarce, as the system is just being introduced to the US market and is initially targeted at centers with high patient volumes ( healthtechhotspot.com, TechBuzz.ai).
Digital Lung Twins
In parallel with AI ultrasound, L&T Technology Services, together with NVIDIA, is developing an AI-powered digital twin platform for the lungs, which will also be presented at RSNA 2025 ( Business Wire, ltts.com).
This solution integrates with CT imaging and uses deep learning models for segmentation to create a patient-specific 3D model of the lung. This model represents the bronchial tree, vessels, lung lobes, and lesions ( Business Wire). ). Technically, the platform is based on NVIDIA MONAI for medical image segmentation and NVIDIA TensorRT for accelerated inference. This combination is intended to provide models with low latency and clinical image quality ( Business Wire).
Source: pl.linkedin.com
Close-up of the Samsung R20 ultrasound device.
The resulting “bio-digital twin” of the lung offers interactive visualization, precise path planning, and navigation support for bronchoscopy. This is useful in planning how to guide a bronchoscope to a peripheral nodule that appears only as a small shadow on a 2D CT ( Business Wire). ). L&T emphasizes that static CT snapshots are to become “living” models that change with the course of the disease, enabling dynamic progression simulation for lung tumors, COPD, and infectious diseases ( Business Wire).
This project builds on a development that has been visible in scientific literature for some time: organ-specific digital twins for the heart and lungs are increasingly discussed for precise simulations and personalized therapy planning ( Nature, MDPI, PMC). ). Review articles on medical digital twins emphasize three building blocks: a physical twin (patient), a virtual twin (model), and a bidirectional connection through which real measurement data updates the model and simulation results feed into real decisions ( Nature).
For the lungs, this specifically means: CT data, functional diagnostic measurements, and in the future possibly wearables or ventilation parameters are fed into the model. This can then be used to simulate flow conditions, ventilation, load on individual lobes, or the impact of planned resections ( Nature, PMC).
AI in Radiology
For ultrasound, there is already concrete data on how significantly AI-assisted imaging can influence the radiological workflow. Studies were presented at RSNA in which AI assistance in musculoskeletal sonography reduced reporting time by approximately 31 % reduzierte, compared to conventional workflows, without compromising diagnostic quality. Such figures are relevant because they directly address the challenges of everyday practice: overcrowded worklists, night shifts with many cases, and the compromise between thoroughness and throughput ( Nature).
A typical emergency room scenario requires multiple eFAST examinations, a POCUS for dyspnea, and an abdominal ultrasound for upper abdominal pain simultaneously. AI-assisted systems can automatically perform relevant standard measurements in such situations, file image series sorted into predefined protocol layouts, and perform plausibility checks during the scan before the images appear in the PACS ( healthtechhotspot.com, Nature).
The role of radiologists thus shifts away from manual routine tasks towards quality control, correlation with clinical findings, and decision support. These are precisely the functions that make human expertise indispensable in a team ( The Washington Post).
Future of Diagnostics
Digital twin concepts have been established in industry, such as in aerospace or automotive manufacturing, for years to simulate design decisions and detect errors early ( Nature). ). In medicine, the focus initially was on the heart, for example, in the “Living Heart” project, and is now expanding to other organs such as the lungs, brain, and skeletal system ( Nature).
In the context of COVID-19, digital twin approaches for the lungs were already developed to optimize ventilation settings and resource planning, for example, in the “BreathEasy” project ( Nature, PMC). ). What is new about the current initiatives is that with platforms like those from L&T and NVIDIA, product-ready solutions for clinical practice are to be created, including integration into image archiving systems, bronchoscope navigation, and prospectively also tumor boards ( Business Wire).
Source: news.samsung.com
Modern Samsung ultrasound devices with AI integration.
Concrete use cases for digital twin technology in medicine are diverse: a chest CT of a COPD patient can be converted into a simulation model to explore the consequences of volume reduction or a resective surgery. For a suspicious nodule, bronchoscopy planning can be done in the virtual twin before the operation begins ( Business Wire, PMC).
Like any new tool, this technology also carries risks and open questions, including validation across different scanner types and patient groups, regulatory classification as a medical device, liability issues in case of navigation errors, or the transparency of the underlying models ( Nature). ). At the same time, the approach opens up the possibility of supplementing clinical trials “in silico”, personalizing therapies better, and virtually testing complex interventions in advance ( Nature, Nature).
The combination of AI ultrasound like the Samsung R20 and organ-specific digital lung twins shows that AI in diagnostics is moving out of the experimental phase and is increasingly being integrated into the infrastructure of radiology and functional diagnostics ( TechBuzz.ai, Business Wire). ). Radiology teams are reacting to this not because “AI is trendy”, but because staff shortages, increasing case numbers, and increasingly complex image data are difficult to manage without automation ( Nature, The Guardian).
The experiences with the Samsung R20 ultrasound will need to show in the coming months whether the AI tools and ergonomic optimization actually lead to less overload, a lower error rate, and higher examination quality in clinical practice, or whether additional complexity arises from configuration and false alarms ( healthtechhotspot.com, TechBuzz.ai). ). The same applies to NVIDIA Digital Twin lung diagnostics: the practical added value will be measured by whether bronchoscopies become safer, resections more plannable, and COPD courses more predictable, and whether this benefit can justify the additional effort for modeling and data integration ( Business Wire, PMC).
It is clear that AI-assisted imaging and digital twin platforms will not replace radiologists, pulmonologists, and interventional teams, but they will force them to redefine their roles. Moving away from the “click workstation” towards orchestrating algorithms, clinical data, and patient needs ( The Washington Post, Nature). ). In this role, AI can take over routine tasks, sort complexity, and thus strengthen the human part of medicine – communication, prioritization, responsibility – rather than pushing it aside ( Nature).