A 98,229-family medtech AI landscape reveals where healthcare AI is actually being patented, who owns it, and where white space remains.
Every quarter, someone publishes a “top 10 medtech trends” article that reads like the last one with the dates changed. Surgical robotics. Medical imaging AI. Wearable sensors. Longevity therapeutics. Yes, yes, yes, yes. What the listicles never answer is the question that actually matters if you’re building, investing, or advising in this space: where is the real innovation happening right now, and who is doing it?
So we pulled the data.
Using the Minesoft Origin API, we indexed every patent published in the US or at the European Patent Office covering surgical robotics (CPC A61B34), healthcare informatics (G16H), surgical accessories and AR/VR guidance (A61B90), diagnostic devices (A61B5), and medical image analysis (G06T7) from January 1, 2025 through April 20, 2026.
The dataset: 98,229 patents. 63,064 US filings (64.2%) and 35,165 EP filings (35.8%).
Here’s what it actually tells us.
If you look at the CPC breakdown, the picture is cleaner than the narrative suggests:
Two things jump out.
First, computer vision is now the dominant software layer in healthcare. Add G06T and G06V and you get 62,316 medtech-tagged filings in 15 months. That is the largest single technology wave in healthcare innovation right now, and it’s not the one the press covers. Everyone writes about LLM-based clinical assistants. The patents are going into image processing.
Second, A61B34 (surgical robotics) is a small but strategic subset. It’s where the moat-builders are concentrating. More on that below.
Here are the top filers across all 98,229 medtech patents:
Three surprises here.
Samsung is the #1 medtech patent filer right now. Not Philips, not Medtronic, not Intuitive. Samsung. They’re building a platform across imaging, wearables, and device integration, and they’re filing like they intend to compete with the entire legacy medtech stack. Anyone who writes Samsung off as “just consumer electronics” in healthcare is looking at the wrong window.
Apple (667 filings) outfiles Stryker, Medtronic, and most legacy device makers. Apple Watch health claims get the coverage, but the patent portfolio tells you Apple is going deep into diagnostic-grade sensing and algorithmic health interpretation. Their legal exposure surface, and their competitive threat to traditional medtech, is expanding fast.
Two J&J subsidiaries (Biosense Webster and Cilag) make the top 10. J&J’s total medtech filing volume, if you add their subsidiaries, rivals Samsung and Philips. The J&J split structure masks the portfolio depth.
Of the 98,229 total medtech patents, only 12,105 explicitly invoke AI, machine learning, deep learning, or neural networks in the title, abstract, or claims. That’s 12.3%.
Read that again. The 80/20 of medtech patents are still classical algorithms, mechanical systems, chemistry, and sensor physics. AI is a minority filing lens, not a majority one.
But inside the AI subset, the rankings shift hard:
Philips almost doubles Siemens in AI-specific medtech filings. Whatever the analyst reports say about Siemens leading in imaging AI, Philips is outbuilding them on raw patent output. That matters for downstream licensing and cross-license posture.
The academic and startup surprises. UC San Diego ranks #6 with 96 AI-medtech patents in 15 months. That’s a larger filing pace than Medtronic. Anumana, a cardiac AI startup, at 72 filings is a volume you don’t expect from a private company. These are worth tracking.
Samsung dropped out of the top 15 in the AI subset. They file on wearable hardware, sensors, and device architecture, not on AI methods claimed explicitly. That is a deliberate strategy, and the opposite of Philips’s approach.
Headline rankings are interesting. Strategic postures are more useful. We cross-tabbed the top 10 medtech filers’ 2025-2026 portfolios against their CPC subclass distribution to see where their patent dollars actually go. The results sort into five distinct archetypes.
AI-first, no device moat. NEC files 1,235 patents in G06T (image processing) and 890 in G06V (computer vision) against only 378 in A61B (devices). Canon is even more extreme: 1,550 G06T and 668 H04N (video signal), with just 168 A61B patents. These are software and imaging companies selling algorithms into other people’s scanners. They are not building device businesses.
Device-first, no AI. Covidien (1,019 of 1,030 patents in A61B) and Cilag (1,285 of 1,317 in A61B) are near-pureplay surgical device filers. Biosense Webster is the same story for cardiac catheters: 1,864 A61B and 430 A61M (drug/fluid delivery). These companies have essentially no G06N (explicit AI method claims) and modest G16H (informatics). The moat is the device, not the software.
Modality-integrated imaging hybrids. Siemens Healthineers and Fujifilm both file heavily in A61B AND G06T – Siemens at 887/649, Fujifilm at 1,171/1,170. Siemens adds 355 G01R (MRI coil design) that Fujifilm doesn’t. These are companies selling hardware AND the AI that runs on their hardware. Vertically integrated.
Full-stack platform. Philips is the only top-10 filer with real depth across all four layers: 2,222 A61B (devices), 1,340 G16H (informatics), 1,115 G06T (imaging AI), and 294 G06N (explicit AI methods). Nobody else has four CPCs above 1,000. This is the platform posture, and it is why Philips shows up #1 in the AI subset ranking – they are the only top-10 filer investing across the full medtech stack simultaneously.
Robotics-first. Intuitive Surgical is the only top-10 filer with significant B25J (general-purpose robotics): 306 patents, on top of 1,459 A61B. Nobody else in this peer set has more than 58 B25J filings. Intuitive’s moat is mechanical and kinematic, not algorithmic.
And then there is Samsung, which does not fit any of these archetypes. Samsung files 2,537 G06T, 1,308 G06V, 1,133 G06F, and 1,111 A61B. Four CPCs above 1,000, but the mix is the opposite of Philips. Samsung is importing consumer-electronics IP strategy into medtech: lots of imaging AI, lots of general computing, devices as table stakes. They are not building a medtech company. They are adding medtech to an existing platform.
The punchline: if you know a top-10 filer’s #1 CPC, you know their strategy. A61B #1 means device moat. G06T #1 means algorithm moat. G16H #1 would mean workflow moat, but nobody’s there yet. That gap is the interesting one.
Zooming in on the last six weeks makes the intervention concentration even sharper. We ran an LLM classification on the 100 most recent medtech filings from these same ten companies (March 10 – April 20, 2026) against seven tech categories. 82% of those recent filings landed in surgical intervention: 51% in Image-Guided Intervention (navigation, AR, registration, catheter visualization) and 31% in Surgical Robotics. Imaging AI, diagnostic sensing, informatics, and patient monitoring together accounted for 7%. Intuitive alone drove 36 of the 109 YES classifications; Cilag, Biosense, and Covidien combined for another 50. Whatever these companies are racing to protect right now, it is inside the operating room, not around it.
The chart also exposes each company’s strategic lane. Intuitive is the only filer with meaningful depth in BOTH intervention categories (19 Image-Guided + 16 Surgical Robotics). Cilag is a pure instruments play (9 of 13 in Surgical Robotics). Biosense is catheter-centric, splitting across image-guided and therapeutic delivery. Philips and Siemens are diversified but thin at this cadence. And the imaging/AI-heavy filers from the full-year stats – Samsung, NEC, Canon, Fujifilm – barely register here, confirming their recent filings sit outside the intervention taxonomy entirely.
Longevity is the hottest investor thesis in healthcare. So who actually owns the IP?
We ran a separate search: patents mentioning longevity, aging, senescence, senolytics, or anti-aging therapeutics, cross-referenced with therapeutics or drug CPC codes, US+EP, since January 2025. Result: 3,153 patents.
Top filers:
Three things to notice.
Academia dominates. UC San Diego, INSERM, CNRS, Harvard, Johns Hopkins, Columbia. Five of the top ten are academic institutions. This is not a normal pharma patent distribution. Longevity IP is still upstream, still in universities, still not consolidated into big pharma portfolios.
Nestle is #3. Nestle Health Science has been quietly building a longevity-adjacent nutritional therapeutics portfolio for years. Amazentis (#5) is Nestle’s partnered longevity spin-out. Combined, “Nestle world” is the largest single longevity filer, outranking every biotech on the list.
Big pharma is under-indexed. Novartis, Regeneron, Bristol Myers Squibb, Genentech show up, but none are in the top five. Either big pharma does not believe longevity is IP-defensible yet, or they’re buying rather than building. Our read: it’s both.
The data does not point to one generic medtech AI strategy. It points to different patent postures depending on the lane.
Surgical robotics is a competition with Intuitive and J&J. The strongest openings are not generic robot mechanics. They are workflow, data handoff, pre-operative planning, post-operative integration, and the software layer around the operating room.
Medical imaging AI has a visible incumbent in Philips. A smaller company needs vertical specificity: one modality, one disease state, one clinical decision, and a filing strategy tight enough to survive comparison against the incumbent stack.
Wearables and diagnostic sensing are increasingly platform fights. Apple owns hardware and ecosystem leverage. The available white space is clinical context, algorithmic interpretation, validation, and closed-loop use cases.
Longevity therapeutics still looks upstream. Universities and research institutions hold much of the early patent position. Licensing and sponsored-research strategy may matter as much as internal filing velocity.
A lot of the founders I talk to in medtech treat patent strategy as a task, not a system. File when the attorney says it’s time. Skip when the budget’s tight. Audit once a year if at all.
The companies on the top-filer list don’t operate that way. Samsung, Philips, Intuitive, and Apple have patent strategy groups that sit adjacent to R&D, product, and regulatory, not downstream of them. Every major invention gets a filing decision within weeks, not quarters.
You won’t match their filing volume. You can match their cadence, their discipline, and their strategic clarity. That’s what a good IP consultant buys you. Not more filings. Better ones, at the right moments, aimed at the right defensive and offensive posture.
Dataset: All patents with publication date on or after January 1, 2025, through April 20, 2026, assigned to country codes US or EP, with at least one CPC classification in A61B34, G16H, A61B90, A61B5, or G06T7. Retrieved via Minesoft Origin API.
AI subset: The same dataset filtered for title, abstract, or claims text matching “artificial intelligence,” “machine learning,” “deep learning,” or “neural network” within standard proximity operators.
Longevity subset: Independent search on CPC classes A61, C12N, or G16H with title/abstract/claims text matching longevity, aging, senescence, senolytic, anti-aging, or age-related, combined with therapeutic, treatment, drug, or method terms.
Assignee normalization: Minesoft probable-assignee field, which consolidates corporate variations (e.g., Philips Koninklijke NV, Koninklijke Philips NV, Philips Electronics = Koninklijke Philips NV).
Top-10 CPC cross-tab: For the archetype analysis, we ran a separate two-dimensional statistics query on the top 10 medtech filers covering all of 2025 and 2026 YTD (21,326 patents), cross-tabulating assignees_name_probable against cpcfirst4chars to produce each filer’s CPC distribution. CPC subclasses referenced: A61B (devices), A61M (drug delivery), A61N (electrotherapy), B25J (robotics), G01R (electric/magnetic measurement), G05B (control systems), G06F (computing), G06N (AI/ML methods), G06T (image processing), G06V (computer vision), G16H (healthcare informatics), H04N (picture/video signal).
Six-week classification: For the intervention concentration finding, we took the 100 most recent filings from the same top-10 assignees (March 10 – April 20, 2026) and ran binary LLM classification against seven medtech categories (Surgical Robotics, Medical Imaging AI, Diagnostic Sensing, Clinical Decision Support, Image-Guided Intervention, Patient Monitoring, Therapeutic Delivery). Haiku 4.5 for the primary pass, Sonnet 4.5 escalation on ambiguous cases. Patents can be YES in multiple categories. Classification records are retained in the project workspace.
Need to understand what this means for your portfolio? ipCapital Group helps leadership teams turn patent landscapes into practical decisions about filing strategy, competitive diligence, M&A readiness, and monetization.
Start a conversation with ipCapital Group or download the IP Strategy Playbook.
Work with ipCapital Group
From invention to monetization, our team has guided 2,000+ engagements across the full IP lifecycle. Start with a free 30-minute discovery call.
Written by
Seth Cronin