Software as a Medical Device: How AI-Powered Software Is Becoming the Medical Device

From stroke detection to cancer screening, SaMD is redefining what a medical device can be – and it doesn't need a single piece of hardware to do it.

Your next prescription might not come in a bottle. It might be a download.

Software as a Medical Device – SaMD – is one of the fastest-growing categories in healthcare. These aren't wellness apps that count your steps. They're clinically validated, FDA-regulated programs that function as standalone medical devices, running on smartphones, laptops, and cloud servers. No factory line. No shipping container. Just code that saves lives.

The numbers reflect how far this has come. In 2025, the FDA cleared roughly 300 AI-enabled medical devices, and 62% were SaMD. More than 1,250 AI-enabled devices are now authorised for marketing in the United States. The shift from hardware to software in healthcare is well past proof-of-concept.

SaMD vs. SiMD: Why the Distinction Matters

The International Medical Device Regulators Forum defines SaMD as "software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device."

SaMD operates independently. An AI algorithm that analyses CT scans for stroke. A mobile app that adjusts insulin recommendations. The software itself is the device.

This differs from Software in a Medical Device (SiMD) – embedded software that powers hardware, like the firmware inside a pacemaker. The clinical function belongs to the hardware; the software is the mechanism.

The distinction matters because SaMD can be deployed globally with a software update, scaled without manufacturing constraints, and updated with a PCCP-approved change rather than a full regulatory resubmission.

Where SaMD Is Saving Lives Right Now

The most mature applications are in medical imaging, which accounts for over 71% of all AI/ML device clearances in the US. Beyond imaging, SaMD is expanding rapidly across cardiology, oncology, mental health, and chronic disease management.

Stroke detection.Viz.ai's AI platform is deployed across more than 1,600 hospitals, with multiple FDA-cleared algorithms spanning neurovascular, cardiovascular, vascular, and trauma applications. In a multicenter study presented at the 2025 International Stroke Conference, it reduced stroke treatment time by an average of 31 minutes. Clinical data also show a 40% reduction in disability at 90 days. In stroke care, each minute of delay in endovascular therapy translates into several days of disability-adjusted life lost.

Sepsis prediction. Prenosis received FDA De Novo marketing authorisation in 2024 for Sepsis ImmunoScore – an AI tool that analyses 22 parameters combining biomarkers and clinical data to assess a patient's sepsis risk within 24 hours. It integrates directly with electronic health records, giving clinicians an earlier warning than traditional clinical observation allows.

Diabetic retinopathy screening. IDx-DR, now called LumineticsCore, made history in 2018 as the first fully autonomous AI diagnostic cleared by the FDA. It screens for diabetic retinopathy without a specialist present, with sensitivity and specificity in the high-80s to low-90s, bringing specialist-level eye screening into primary care offices for the first time.

Digital therapeutics. Akili Interactive's EndeavorRx was the first FDA-cleared video game therapy for paediatric ADHD in 2020. Programmes targeting substance use disorder and diabetes management continue to expand the category, with randomised controlled trial data supporting their use alongside traditional treatment.

The AI Acceleration: Foundation Models Enter the Clinic

The next frontier in SaMD is foundation models – large AI systems trained on massive datasets that apply learned patterns across multiple clinical applications, rather than being trained separately for each condition.

Aidoc, whose aiOS platform holds 18 FDA clearances covering conditions from stroke to pulmonary embolism and cervical fractures, received funding in 2025 to develop its CARE (Clinical AI Research Engine) foundation model. The goal is a single learning system that spans multiple clinical domains.

Tempus acquired digital pathology AI company Paige in August 2025, with CEO Eric Lefkofsky describing the deal as substantially accelerating their effort to build the largest foundation model in oncology. Tempus combines one of the world's largest clinical and molecular datasets with AI analytics for precision medicine.

The shift toward foundation models changes the economics and speed of SaMD development. A well-trained foundation model can be adapted to a new clinical use case far faster than training a condition-specific algorithm from scratch – which means more applications, faster, at lower cost.

The Regulatory Game Changer: Predetermined Change Control Plans

Traditional medical device regulation required a new submission for every significant software update. For AI-based SaMD that improves over time, that created a structural problem: the regulatory framework was built for static devices.

Predetermined Change Control Plans (PCCPs) are changing that.

In October 2023, the FDA, Health Canada, and the UK's MHRA jointly published five guiding principles for PCCPs in machine learning-enabled medical devices. By early 2025, the FDA finalised detailed PCCP guidance for AI-enabled device software functions.

A PCCP lets manufacturers map out anticipated software modifications at the time of their initial regulatory submission. If the plan is authorised, they can implement pre-specified updates – including AI model retraining and performance improvements – without filing a new marketing application for each change. Each plan requires three core elements: a description of planned modifications, a modification protocol setting out validation steps, and an impact assessment analysing how changes affect safety.

The tri-regulator alignment between the US, UK, and Canada gives manufacturers greater confidence that adaptive AI models have a credible regulatory path in major markets.

Global Companies to Watch

Lunit (Seoul, South Korea) serves more than 10,000 healthcare providers across 65+ countries. Its INSIGHT suite detects chest abnormalities with 97–99% accuracy and breast cancer with 96% accuracy. In 2024, Lunit acquired New Zealand-based Volpara Health Technologies to create a comprehensive cancer intelligence portfolio. Its AI was selected for Australia's national breast cancer screening programme – widely described as a global first.

Huma Therapeutics (London, UK) received FDA Class II clearance in 2023 for a disease-agnostic SaMD platform applicable across multiple conditions. The platform integrates with external devices including heart rate monitors and glucose meters, and hosts AI algorithms supporting screening, diagnosis, and clinical decision-making.

XUND (Vienna, Austria) is an MDR-certified SaMD platform for digital triage and diagnosis, serving more than 10 million patients. XUND raised funding in early 2025 to expand across the DACH region and the UK.

[H2] The Challenges That Remain

Regulatory fragmentation. Multi-region SaMD launches can cost several hundred thousand to multiple millions of dollars. The EU's MDR often classifies software at higher risk levels than US guidelines, and global harmonisation is still a work in progress. For companies targeting international markets, navigating three or more regulatory frameworks simultaneously adds significant time and cost.

Algorithmic bias. To date, the FDA has only approved AI/ML tools using locked algorithms that don't change after deployment. The PCCP framework is opening the door for adaptive models, but training data diversity remains a critical challenge. An algorithm trained on data from one patient population may perform less accurately for patients not well represented in the training set.

Reimbursement uncertainty. Even with FDA clearance, insurance coverage for SaMD is inconsistent. Germany's DiGA pathway, which created direct reimbursement for approved digital therapeutics, is frequently cited as a model other markets could follow. Without reimbursement, even clinically proven SaMD faces a commercial ceiling.

FDA capacity constraints. Double-digit workforce reductions at the FDA in 2025 have added another layer of risk for developers, even as AI-enabled submissions continue to surge.

What SaMD cannot yet do. Adaptive AI in clinical settings remains the exception rather than the rule. Reimbursement frameworks outside Germany lag behind clinical evidence. Algorithmic bias in underrepresented populations is a known, unresolved risk. These are not reasons to avoid the category – they are the due diligence questions every health system should be asking before deployment.

What Comes Next

Software as a medical device represents something genuinely unusual in healthcare technology: a category where the marginal cost of deployment approaches zero, updates can reach millions of patients simultaneously, and clinical evidence is accumulating faster than reimbursement and regulatory frameworks can keep pace.

Algorithms are already detecting strokes faster than radiologists, predicting sepsis hours before symptoms surface, and screening for eye disease in clinics that have never had a specialist on site. Patients who would have waited weeks for a diagnosis are getting answers in minutes.

For health system leaders, the question has shifted. It's no longer whether SaMD belongs in your procurement strategy. It's which tools have the clinical evidence, regulatory clearance, and reimbursement pathway to deploy safely – and how to build the internal infrastructure to evaluate them rigorously.

The next life-saving medical device may already be running on a server somewhere. It just needs the right health system to put it to work.

FAQ: Software as a Medical Device

What is software as a medical device (SaMD)?

SaMD is software that performs a medical purpose independently, without being part of a hardware medical device. Examples include AI algorithms that detect strokes in CT scans, apps that predict sepsis risk from electronic health record data, and autonomous screening tools for diabetic retinopathy. It runs on standard computing hardware: smartphones, laptops, and cloud servers.

Does SaMD require FDA approval?

The regulatory pathway depends on the risk level of the intended use. Low-risk SaMD may qualify for exemption or notification pathways. Higher-risk software that directly informs clinical decisions typically requires FDA De Novo authorisation or a 510(k) clearance. In Europe, the MDR applies and often classifies SaMD at higher risk levels than US guidance.

What is the difference between SaMD and a medical app?

A medical app can be a consumer wellness tool – a step counter, a sleep tracker – that doesn't diagnose or treat. SaMD has a specific medical purpose, is clinically validated, and is regulated as a medical device. The defining question is whether the software is intended to diagnose, treat, mitigate, or prevent a disease or condition.

What is a Predetermined Change Control Plan (PCCP)?

A PCCP lets AI-based SaMD developers pre-specify anticipated software changes at the time of initial authorisation. If approved, developers can implement those changes without filing a new marketing application for each update. The FDA finalised PCCP guidance in early 2025, endorsed jointly with Health Canada and the UK's MHRA.

What are the main barriers to SaMD adoption?

Regulatory fragmentation across global markets, inconsistent reimbursement (particularly outside Germany), algorithmic bias in underrepresented patient populations, and integration with existing electronic health record infrastructure. Cost and implementation complexity also affect procurement decisions in resource-constrained health systems.

Which clinical areas are most advanced in SaMD?

Medical imaging leads, accounting for over 71% of all AI/ML device clearances in the US. Radiology applications for stroke, pulmonary embolism, and fracture detection are well established. Cardiology, oncology, sepsis prediction, and digital therapeutics for ADHD and substance use disorder represent the fastest-growing adjacent areas.

Stay ahead of what’s next in healthcare.

Healthy Innovations is my weekly newsletter delivering strategic analysis of emerging biotech and digital health.

No spam. Unsubscribe anytime.