Artificial intelligence and machine learning (AI/ML) are increasingly being integrated into medical systems. This delivers smarter and faster care by bringing intelligence closer to where the data is created and used.
The most recent advances in computer vision, large language models, edge computing, and real-time signal processing are improving medical diagnosis and reducing the latency between data acquisition and actionable clinical results. This is enhancing medical imaging use cases and enabling advances in robotic surgery and remote monitoring while delivering more integrated systems to improve patient care.
We introduce some real-world products and solutions that demonstrate how the combination of AI/ML and healthcare has become a reality.
Medical imaging
In medical imaging, AI/ML techniques, including convolutional neural networks, are being deployed directly within magnetic resonance imaging (MRI) and computed tomography (CT) systems. Platforms such as Royal Philips’s AI-enhanced MRI and CT suites and Aidoc’s radiology triage are examples of how models trained using large datasets, helping to detect critical conditions with very low latency. Butterfly Network has further expanded this concept by embedding AI inference into portable, handheld ultrasound hardware.
Philips’s AI-enhanced MRI and CT
At the 2025 European Congress of Radiology, Royal Philips announced a generation of AI-integrated MRI solutions. The main innovation is SmartSpeed Precise, a system powered by dual AI engines built around the company’s existing Compressed Sense and SmartSpeed platforms.
SmartSpeed Precise improves image sharpness by 80%, enabling a better visualization of anatomical structures. Applied to standard Sense imaging, the technology allows scans to be completed in one-third of the time without affecting the quality of the diagnostic images.
A faster scan time means a more comfortable experience for the patient, with less time spent without moving inside the scanner. Moreover, patients can access a diagnosis more quickly, as wait times are reduced. One clinical site reduced exam times to under 60 minutes per slot for whole-body multiparametric exams, enabling it to scan two additional patients per day.
Recently, Royal Philips also received FDA 510(k) clearance for its Verida system, a next-generation spectral CT scanner that integrates AI-driven reconstruction with a specialized dual-layer detector architecture. At the core of the Verida system is a third-generation Nano-panel Precise dual-layer detector.
Unlike photon-counting detectors that utilize direct conversion, this scintillator-based stack employs two layers (the Nano-panel Precise detector) to capture low- and high-energy photon spectra from a single X-ray source, providing spectral results 100% of the time. This “always-on” technology enables spectral imaging capabilities to be active for every patient, on every scan, without requiring special procedures or separate, time-consuming scans.
Inside the signal-processing chain, the system uses a deep-learning reconstruction engine, a properly trained neural network that provides an estimated 80% reduction of the image noise, maintaining the spatial resolution. The computational back end can handle high-throughput processing, up to 145 images per second, enabling full-volume spectral data processing in under 30 seconds, 2× faster than previously available technology.
Aidoc’s AI-powered clinical platform
Aidoc’s AI-powered platform, adopted by over 1,600 hospitals worldwide, is built around the idea that connected teams deliver better outcomes. At the heart of the solution is aiOS, the company’s proprietary enterprise platform, which operates as an always-on intelligence layer across a health system. The platform covers 75% of the patient population, according to the company, enabling physicians to make accurate decisions using real-time data and allowing care teams in multiple departments to collaborate on a unified patient journey.
The clinical solutions have five core specialties. Radiology solutions include image-based triage and quantification, powered by 18 FDA-cleared algorithms and eight FDA-cleared partner algorithms. Beyond imaging, Aidoc extends into cardiology, neurovascular care (including stroke and brain aneurysm detection), vascular conditions such as pulmonary embolism and aortic disease, and spine solutions.
A mobile care coordination app delivers real-time notifications for time-sensitive cases, built-in risk stratification, and a mobile imaging viewer, with electronic health-record data automatically fed in to facilitate communication between divisions. Adopted by several leading health systems, Aidoc has delivered significant improvements, including turnaround-time reductions of up to 55% for intracranial hemorrhage cases and length-of-stay reductions of up to 26% for pulmonary embolism cases.
Aidoc’s AI-powered triage platform for large vessel occlusions and medium vessel occlusions proved effective in a study presented at the International Stroke Conference 2026.
In a comparative study of 1,557 CT angiography exams by the University of Texas Medical Branch, Aidoc also showed 92.6% sensitivity for large vessel occlusions, a rate significantly higher than the 70.4% offered by traditional solutions.
Butterfly Network’s ultrasound technology
Butterfly Network, a company specializing in ultrasound technology, has received FDA 510(k) clearance for its Gestational Age (GA) Tool, the first “blind-sweep” AI-powered ultrasound application for pregnancy dating. The technical innovation consists of replacing traditional piezoelectric transducer arrays with Ultrasound-on-Chip technology, which integrates a complete ultrasound system onto a single CMOS chip.
The GA Tool employs a deep-learning inference engine trained with a dataset of >21 million ultrasound images. In contrast to conventional fetal biometry that requires precise manual alignment and measurement of the biparietal diameter or femur length by an experienced sonographer, the “blind-sweep” method employs a simplified acquisition protocol. The operator performs guided probe sweeps over the maternal abdomen without the need to interpret images in real time or optimize for targets.
The AI algorithm then examines the resulting volumetric data to estimate the gestational age between 16 and 37 weeks. The system learns the mapping of anatomical landmarks to gestational maturity to ensure high fidelity of diagnosis and provides results similar to traditional biometric assessments in less than two minutes.
Robotic surgery and remote monitoring
In the field of surgery, robotic platforms such as Intuitive Surgical Operations Inc.’s da Vinci system include real-time haptic feedback loops and computer-assisted motion scaling to minimize the distance between the surgeon’s input and the end-effector’s output. Edge ML models on devices such as DexCom Inc.’s G7 and Eko Health Inc.’s cardiac sensors analyze continuous streams of biosignals locally, sending only clinically relevant anomalies, as in remote monitoring.
Intuitive’s da Vinci 5
It is one of the most advanced robotic systems for minimally invasive surgery. The name is no coincidence. Leonardo da Vinci was the first to study the movements of the human body systematically and, in 1495, to design a prototype of a humanoid robot, the mechanical knight (called automa cavaliere in Italian).
With a processing capacity 10,000× greater than the previous Xi generation, the da Vinci 5 supports a suite of digital and tactile features intended to improve surgical precision through real-time data analysis and haptic integration.
A major technical upgrade is the introduction of Force Feedback technology. It uses force-sensing instruments and new algorithms to measure the physical resistance that the robotic arms encounter. This information is then fed back to the surgeon’s console so that the surgeon can “feel” tissue tension and resistance. To complement this haptic information, the system also incorporates a Force Gauge, a real-time visual display of pressure information.
The Intuitive Hub and its ML models provide the AI capabilities of the system. These models produce automated case insights by algorithmically dividing surgical video into discrete phases such as dissection, retraction, and suturing. The system provides objective performance metrics based on the instrument kinematics and phase durations. These metrics can be used to compare individual surgical techniques against standardized clinical benchmarks, with a focus on motion efficiency and instrument economy.
During active procedures, the da Vinci 5 utilizes predictive analytics and vision-based AI to enhance the surgical field. The in-console video replay feature allows surgeons to access recorded segments of the ongoing procedure directly through the 3D viewer. This function is supported by AI overlays that can highlight specific anatomical landmarks or track instrument paths.
Dexcom G7
The Dexcom G7, a continuous glucose monitor (CGM), has evolved into an AI-driven health device. According to recent updates from DexCom, the G7 (Figure 3) now integrates sophisticated AI to simplify daily diabetes management and deliver deeper metabolic insights.
A relevant AI feature is Smart Food Logging. This tool uses computer vision and ML to analyze photos of meals taken within the app. The AI automatically identifies ingredients and generates meal descriptions, significantly reducing the manual effort required for carb counting and data entry.
Furthermore, DexCom has launched a proprietary generative AI platform powered by Google Cloud’s Vertex AI and Gemini models. This platform analyzes individual health data patterns to offer personalized “Weekly Insights”—that is, recommendations based on the user’s glucose trends, activity levels, and sleep patterns.
Eko Health Sensora platform
Eko Health has achieved a major milestone with the FDA clearance of its Eko Foundation Analysis Software with Transformers (EFAST) algorithm, the first-ever cardiac “foundation model” designed for clinical use. This AI has been integrated into the Sensora platform, Eko’s digital stethoscope, which helps clinicians detect signs of potential cardiac diseases quickly and accurately (Figure 4).
The EFAST algorithm integrates several advanced AI features that transform the diagnostic process. Instead of being trained for a single purpose, the system utilizes a cardiac foundation-model architecture trained on over 4 million heart-sound and electrocardiogram recordings. This large dataset allows the AI to learn a general representation of cardiac health, which is then fine-tuned to detect specific conditions, such as structural heart murmurs and atrial fibrillation, with high specificity.
The post AI/ML integration in medical systems appeared first on EDN.