Executive Summary 🔗
- FDA authorized 86 AI/ML devices in Q2 2026: 84 510(k)s and 2 De Novos.
- Decisions ran from April through June 2026. June was the heaviest month, with 28 authorizations.
- Sixty-three records were SaMD. The remaining 23 show the other side of the AI device market: AI embedded in imaging systems, sensors, procedure tools, diagnostic hardware, or acquisition/reconstruction platforms.
- Radiology still dominated with 59 records, but the quarter also included cardiovascular, neurology, GI/urology, anesthesiology, orthopedics, dental, surgery, hematology, pathology, clinical chemistry, and microbiology records.
- Twelve authorization records referenced a PCCP. Those were not confined to radiology: they appeared in anesthesiology, cardiovascular, pathology, and radiology contexts.
- Observed received-to-decision timing varied sharply by route, making route strategy a runway issue as well as a regulatory issue.
- The two De Novos each created a new device category. Automated Imaging Diagnostics'
neuropacscreated a Parkinsonian-syndrome diffusion-MRI classification category, and SpectralMD'sDeepView AI Systemcreated a burn-wound AI aid category tied to healing prediction and tissue segmentation. Every follow-on entrant inherits their special controls.
Most Of The Quarter Ran Through 510(k) 🔗
Eighty-four of 86 records cleared through 510(k). For anyone allocating capital, that ratio is the headline: the predicate-based route, the cheapest and fastest path to market, is absorbing nearly all AI device activity. The use cases, meanwhile, kept spreading.
But the 2 is where the ambition lives. If you want to shoot big and set the rules of a new category, De Novo is the route. The first mover writes the special controls every follow-on competitor inherits. The trade is runway: review runs longer, and the evidence package usually includes studies a predicate path would skip. The advantages are real, and we help teams decide when they justify the extra time and cost.
A note on the numbers. They come from FDA Device Explorer, the database our own analytics team built and maintains. We define AI by industry norms rather than FDA's definition, and our classification updates in real time. FDA's AI list takes months to catch up. If our counts differ from FDA's, that is why.
The cohort stretched from ECG-based amyloidosis screening, sepsis flagging, fetal ultrasound anomaly detection, coronary plaque support, concussion eye tracking, and vascular occlusion triage to breast-pathology prognostics, burn-wound healing assessment, dental disease support, sleep support, radiation-oncology planning, PACS workflow software, and lung-sound wheeze detection. The established radiology core remained large: mammography CAD, colonoscopy CADe, prostate MRI analysis, CT/PET reconstruction, lung nodule support, and ultrasound guidance all appeared in the quarter. The master table below lists all 86 records.
FDA's AI/ML activity is no longer just chest X-ray triage and radiology quantification. If your board is still asking "will FDA clear AI?", that question is settled. The question that sets budget and timeline is "which evidence story fits this intended use, user, environment, and change model?"
There were 72 Traditional 510(k)s, 12 Special 510(k)s, and 2 Direct De Novo records. The Special 510(k)s are the clearest window into what an AI device business looks like after first clearance. They cover architecture moves, such as Aidoc's shift of head CTA triage toward a Linux-based cloud environment; output and model changes, such as Lunit's mammography update; population expansions, such as Imeka's adolescent-use expansion for diffusion MRI analysis; and workflow or system changes, such as Dexcom's AI-generated glucose insights, DESKi's HeartFocus acquisition guidance, Sonio's fetal-ultrasound workflow software, Fujifilm's Synapse PACS update, and Therapixel's MammoScreen update.
Observed review timing reinforces why route strategy is a financing and roadmap issue, not just a regulatory filing detail. In this cohort, Traditional 510(k)s had a median received-to-decision interval of 170.5 days, Special 510(k)s had a median of 29 days, and the two De Novos took 328 and 485 days. Those numbers should not be treated as future FDA timelines, but they are useful for executive planning: the wrong route assumption can change runway, launch sequencing, and fundraising milestones.
That is a good map of where AI device work actually lives after first clearance: architecture changes, output changes, use-population changes, workflow expansions, system integration, and model refinements.
The Two De Novos Set Boundaries 🔗
The two De Novos will outlast most of the quarter's 510(k)s. Each created a new classification, and its special controls will define what every follow-on device in the category must show.
Automated Imaging Diagnostics received De Novo grant DEN240071 for neuropacs, a software application that analyzes diffusion MRI data from patients 40 and older with Parkinson's disease, multiple system atrophy Parkinsonian variant, or progressive supranuclear palsy. The output is a classification report based on degenerative brain patterns, intended to help neuroradiologists and neurologists differentiate those conditions from Parkinson's disease. The intended use is carefully bounded: it supplements standard neurological assessment and other clinical tests; it is not a stand-alone diagnostic; and clinicians must rule out other causes of Parkinsonism before relying on it.
The record requires clinical performance validation to demonstrate diagnostic accuracy and reproducibility against a clinically relevant reference standard, plus software verification, validation, hazard analysis, and technical descriptions of model inputs and outputs. That is the entry price for every future device in the category. De Novo is a category-creation exercise: FDA is not only clearing one product, it is writing the rulebook the first mover's competitors will have to follow.
SpectralMD received De Novo grant DEN250028 for the DeepView AI System, an AI-enabled burn-wound assessment system. The software analyzes optical images, detects deep partial-thickness and full-thickness burn tissue, highlights regions unlikely to heal within 21 days with conservative treatment alone, and calculates wound area and percent total body surface area. It is an aid to physicians and must be used with clinical assessment. It does not provide definitive excision margins.
This is not a classic software-only SaMD example. It pairs AI software with a clinical imaging system and a treatment-planning-adjacent claim, the combination most relevant to hardware companies adding AI to existing product lines. The evidence expectations are correspondingly specific: standalone performance testing, sensitivity and specificity, segmentation accuracy, subgroup analysis, and validation that users improve diagnostic characterization compared with unaided assessment.
Together, the two De Novos show the current FDA posture well. New AI categories are possible, but the claim needs tight boundaries, the reference standard needs to be defensible, and labeling has to prevent over-reading the output.
Radiology Still Dominated, But The Edges Are Moving 🔗
Radiology accounted for 59 of the 86 records. That is expected. Radiology has the longest AI device precedent stack, clear digital inputs, mature reader-study patterns, and many measurable outputs.
The non-radiology records are the planning signal. The cardiovascular panel included six records: Anumana's ECG-AI Cardiac Amyloidosis algorithm, Conavi's Novasight Hybrid System, Abbott's OPTIS Mobile Next imaging system, Biozen's fingertip blood pressure monitor, iRhythm's ZEUS Platform, and Pathwai's EchoNext. Two more cardiac-focused records (Anumana's WatchMate Software and ADAS 3D cardiac image post-processing) were reviewed under the radiology panel, a reminder that panel assignment follows the device and imaging modality, not just the clinical organ system.
Other panels brought in colonoscopy CADe, sepsis flagging, dental disease support, neurological classification and review software, orthopedic control systems, burn-wound assessment, radiation oncology planning, sleep support, and TytoCare's wheeze-detection software, which FDA assigned to the anesthesiology panel.
It is great to see different review panels increase their proficiency in AI review, and very fruitful to see the industry taking advantage of that and deploying AI in other body systems and disciplines other than just radiology.
PCCP Became A Practical Design Topic 🔗
Twelve authorization records referenced a PCCP: Anumana's ECG amyloidosis algorithm, Siemens Healthineers' syngo Dynamics, Conavi's Novasight Hybrid System, ArteraAI Breast, Biozen's BP1000 monitor, SubtleHD-PET, DeepHealth BAC, SmartAlpha's MSK Go, TytoCare's wheeze-detection software, DESKi's HeartFocus, Brainomix 360 Hyperdensity, and Siemens Healthineers' syngo.CT Coronary Cockpit.
Eleven of the twelve PCCP records were Traditional 510(k)s; only DESKi's HeartFocus paired a PCCP with a Special 510(k). Note, however, a PCCP was mentioned in the Special 510k, but it was not added as part of the Special 510k since that is not allowed. The PCCP already existed in the Traditional, and it appears in the Special they are just making known what changes were done as part of a PCCP already cleared in the original traditional 510(k).
PCCPs are no longer a radiology novelty, nor are they reserved for standalone diagnostic algorithms. They appear wherever a sponsor can define a controlled future-change envelope: what can change, how performance will be verified, and how the sponsor will decide whether a change stays inside the authorized plan.
For teams building AI SaMD, the implication is straightforward: treat PCCP strategy as a design input. A PMA is not for everyone, and it can complicate your submission unnecessarily if you don't have a fully baked plan. However, if done correctly, it can save a lot of money by obviating the need for follow-on specials or traditional 510(k)s. If the product roadmap depends on model updates, new data sources, threshold changes, population expansions, or workflow integration changes, the regulatory strategy needs to describe that future state early enough for engineering and validation to support it.
Real-World Data Was Common, But Uneven 🔗
We flagged 37 records with real-world-evidence-related content. Four were marked as primary clinical evidence: Anumana's ECG amyloidosis algorithm, Bayesian Health's sepsis flagging device, Pathwai's EchoNext, and Alpha Intelligence Manifolds' DeepXray Spina. Of the remaining flagged records, 31 used real-world data as supportive validation, one as supplemental clinical evidence, and one for postmarket surveillance.
The distinction is regulatory, not semantic. "We used clinical data" is not the same as "real-world evidence is the primary support for the claim." Many AI submissions use retrospective clinical cases, external validation datasets, procedure reports, images, scans, or clinical records. The question is what regulatory job the data is doing. The way that I like to conceptualize real-world evidence is in a way that significantly moves the needle on the cost of a study. Meaningful real-world evidence is when you don't have to hire humans to ground truth data. Ground truth already exists from the standard of care record.
In Q2, RWE-style evidence appeared in diverse contexts: ECG amyloidosis, sepsis flagging, appendicitis assessment, mammography, sleep support, prostate MRI, ultrasound, CT/PET enhancement, TAVR planning, lung nodule support, radiation oncology planning, vascular occlusion triage, neurology review, and orthopedic planning. This is consistent with the market moving away from one-off curated datasets and toward evidence packages that connect technical performance to deployment-relevant clinical data.
Route Mix Still Means Predicate Strategy 🔗
The route mix is an executive planning signal. Most records still turned on substantial-equivalence arguments, but the evidence burden changes depending on where the AI function sits. A software-only triage or detection system often lives or dies on standalone performance, reader impact, workflow, and labeling. An imaging system with AI reconstruction may need to show image quality, artifact correction, quantitative consistency, and safety in the context of scanner operation.
For leadership, predicate selection, device architecture, and release roadmap need to be decided together. Otherwise, a commercial feature choice can become a validation burden, a review-timeline problem, or a diligence issue later.
What This Means For Investors, Executives, And Founders 🔗
The quarter suggests five planning questions leadership should answer before freezing the roadmap or committing capital.
First, is the claim boundary financeable and reviewable? The most persuasive records are narrow about who uses the output, what the output means, and what it does not mean.
Second, does the predicate story match the product story leadership is selling? If the roadmap requires a predicate leap, a new population, a different architecture, or a new kind of output, the evidence plan has to account for that well before verification closes out.
Third, is the change-control plan owned by engineering, quality, and regulatory together? PCCP is not a regulatory writing exercise after model development. It is a controlled-change contract between the intended use, model lifecycle, validation methods, and labeling.
Fourth, have you separated FDA evidence from buyer evidence? FDA may accept a specific validation package for clearance. A hospital, payer, strategic partner, acquirer, or investor may still ask whether performance holds at their sites, with their users, in their workflow, and under their monitoring expectations. The best consultants will be able to harmonize both, so one is directly reused by the other.
Fifth, are you studying De Novos even if you plan to file a 510(k)? De Novo special controls often become tomorrow's predicate framework. neuropacs and DeepView AI System are not just two products; they are new examples of FDA’s policy setting on AI.
Master Clearance Table 🔗
The table below lists all 86 AI/ML records in the Q2 2026 FDA Device Explorer cohort.
| Date | FDA ID | Route | Company | Device | Panel | Code | SaMD | PCCP | RWE |
|---|---|---|---|---|---|---|---|---|---|
| 2026-04-03 | DEN240071 | De Novo | Automated Imaging Diagnostics | neuropacs | Neurology | SHO | Yes | No | No |
| 2026-04-03 | K253674 | 510(k) | Stryker | Blueprint Patient-Specific Instrumentation | Orthopedics | PHX | Yes | No | No |
| 2026-04-03 | K260524 | 510(k) | United Imaging | uMI Panvivo (uMI Panvivo LS) | Radiology | KPS | No | No | No |
| 2026-04-03 | K260724 | 510(k) | Magentiq Eye, Ltd. | MAGENTIQ-COLO (ME-APDS) | Gastroenterology / Urology | QNP | Yes | No | No |
| 2026-04-06 | K253288 | 510(k) | Esaote, S.p.A. | 6450 Ultrasound System (MyLabE80) | Radiology | IYN | No | No | No |
| 2026-04-07 | K253801 | 510(k) | Anumana | ECG-AI Cardiac Amyloidosis (CA) 12-Lead Algorithm (1040) | Cardiovascular | SHP | Yes | Yes | Yes |
| 2026-04-09 | K252228 | 510(k) | Checkcells | Seaman Pro/Seaman | Hematology | POV | No | No | No |
| 2026-04-10 | K252190 | 510(k) | Aitewan Biomedical Technology | DeepBT Detector-Plus | Radiology | QKB | Yes | No | Yes |
| 2026-04-10 | K252548 | 510(k) | Siemens Healthineers | AI-Rad Companion Organs RT | Radiology | QKB | Yes | No | Yes |
| 2026-04-10 | K253689 | 510(k) | Siemens Healthineers | syngo Dynamics (VA41F) | Radiology | QIH | Yes | Yes | Yes |
| 2026-04-10 | K253930 | 510(k) | Overjet | Overjet Iris Intelligent Imaging System | Radiology | QIH | Yes | No | No |
| 2026-04-10 | K260082 | 510(k) | 3Shape TRIOS A/S | TRIOS Dx (R1) | Dental | SHQ | Yes | No | No |
| 2026-04-14 | K253502 | 510(k) | GE HealthCare | Critical Care Suite with Enteric Tube Positioning AI Algorithm | Radiology | QIH | Yes | No | No |
| 2026-04-17 | K252237 | 510(k) | Edgecare | EdgeFlow UW20 | Radiology | IYO | No | No | No |
| 2026-04-17 | K252945 | 510(k) | Conavi Medical | Novasight Hybrid System | Cardiovascular | OBJ | No | Yes | No |
| 2026-04-20 | K252332 | 510(k) | Mim Fertility | Folliscan | Radiology | QIH | Yes | No | Yes |
| 2026-04-22 | K253163 | 510(k) | Ever Fortune.Ai, Co., Ltd. | EFAI ERSUITE CT APPENDICITIS ASSESSMENT SYSTEM (APPEN-CT-100) | Radiology | QAS | Yes | No | Yes |
| 2026-04-22 | K253256 | 510(k) | Ottobock | myosmart. (13E522) | Neurology | GXY | No | No | No |
| 2026-04-23 | K260320 | 510(k) | Lunit | Lunit INSIGHT MMG (v1.1.10) | Radiology | QDQ | Yes | No | No |
| 2026-04-23 | K260785 | 510(k) | Dentsply Sirona | DS Core CBCT Anatomy | Radiology | QIH | Yes | No | No |
| 2026-04-24 | K253459 | 510(k) | ABBOTT MEDICAL | OPTIS Mobile Next Imaging System (1014932) | Cardiovascular | NQQ | No | No | Yes |
| 2026-04-27 | K252628 | 510(k) | Wesper | CASSIE | Anesthesiology | MNR | Yes | No | Yes |
| 2026-04-29 | K253682 | 510(k) | Quantib B.V. | DeepHealth ProstateAI | Radiology | QDQ | Yes | No | Yes |
| 2026-04-29 | K254184 | 510(k) | Siemens Healthineers | syngo Application Software | Radiology | QIH | No | No | Yes |
| 2026-04-30 | K250680 | 510(k) | Bayesian Health | Bayesian Health Sepsis Flagging Device | Gastroenterology / Urology | SAK | Yes | No | Yes |
| 2026-04-30 | K252503 | 510(k) | Canon Medical | Intelligent NR | Radiology | OWB | Yes | No | No |
| 2026-04-30 | K254189 | 510(k) | Fujifilm | Synapse 3D Base Tools (V7.2) | Radiology | QIH | Yes | No | No |
| 2026-05-04 | K252954 | 510(k) | Savelife.ai | MammoSightAI | Radiology | QFM | Yes | No | Yes |
| 2026-05-04 | K254115 | 510(k) | Artera | ArteraAI Breast | Pathology | SHW | Yes | Yes | No |
| 2026-05-05 | K252558 | 510(k) | Philips | Lumify Diagnostic Ultrasound System | Radiology | IYN | Yes | No | Yes |
| 2026-05-06 | K253152 | 510(k) | Biozen, LLC | BP1000 Ultra-Compact Fingertip Blood Pressure Monitor (BP1000) | Cardiovascular | DXN | No | Yes | No |
| 2026-05-07 | K253992 | 510(k) | Restor3D | Veritas Reverse Total Shoulder System | Orthopedics | PHX | No | No | No |
| 2026-05-08 | K260419 | 510(k) | Hangzhou ChohoTech Co., Ltd. | LingOral Dental Design System | Dental | PNN | Yes | No | No |
| 2026-05-11 | K252953 | 510(k) | Velmeni | Velmeni for Dentists (V4D) Endo-Perio | Radiology | MYN | Yes | No | No |
| 2026-05-12 | K260378 | 510(k) | AZmed | Rayvolve | Radiology | QBS | Yes | No | No |
| 2026-05-13 | K260032 | 510(k) | BeauBrain Healthcare | Morph | Radiology | QIH | Yes | No | No |
| 2026-05-14 | K254013 | 510(k) | Subtle Medical | SubtleHD-PET (1.x) | Radiology | LLZ | Yes | Yes | Yes |
| 2026-05-14 | K261317 | 510(k) | Aidoc | BriefCase-Triage | Radiology | QAS | Yes | No | Yes |
| 2026-05-15 | K252563 | 510(k) | Neurophet. | Neurophet SCALE PET | Radiology | QIH | Yes | No | No |
| 2026-05-15 | K253716 | 510(k) | United Imaging | uSONIQUE Genesis, uSONIQUE Genesis Pro, uSONIQUE Genesis Elite, uSONIQUE Genesis Super, uSONIQUE Pulse, uSONIQUE Pulse Pro, uSONIQUE Pulse Elite, uSONIQUE Pulse Super, uSONIQUE Venus, uSONIQUE Venus Pro, uSONIQUE Venus Elite, uSONIQUE Venus Super | Radiology | IYN | No | No | Yes |
| 2026-05-15 | K254120 | 510(k) | Subtle Medical | SubtleHD-CT (1.x) | Radiology | QIH | Yes | No | Yes |
| 2026-05-15 | K261273 | 510(k) | Neurophet. | Neurophet AQUA | Radiology | QIH | Yes | No | No |
| 2026-05-20 | K253596 | 510(k) | Canon Medical Systems | Aquilion ONE (TSX-308A/TSX-306A) V2.0 | Radiology | JAK | No | No | No |
| 2026-05-21 | DEN250028 | De Novo | SpectralMD | DeepView AI® System | Surgery | SHY | No | No | No |
| 2026-05-21 | K253720 | 510(k) | United Imaging | uSONIQUE Nova, uSONIQUE Nova Elite, uSONIQUE Vita, uSONIQUE Vita Elite, uSONIQUE Grace, uSONIQUE Grace Elite | Radiology | IYN | No | No | No |
| 2026-05-21 | K254131 | 510(k) | DeepHealth | BAC | Radiology | QIH | Yes | Yes | Yes |
| 2026-05-21 | K260234 | 510(k) | Smart Alfa Teknoloji San. Ve Tic. A.S. | MSK Go | Radiology | QIH | Yes | Yes | No |
| 2026-05-21 | K261359 | 510(k) | Dexcom | Stelo Glucose Biosensor System | Clinical chemistry | SAF | No | No | No |
| 2026-05-22 | K260113 | 510(k) | United Imaging | uMR Astra | Radiology | LNH | No | No | No |
| 2026-05-22 | K260324 | 510(k) | Canon Medical Informatics | Vitrea CT Transcatheter Aortic Valve Replacement (TAVR) Planning | Radiology | QIH | Yes | No | Yes |
| 2026-05-27 | K253971 | 510(k) | Visby Medical | Visby Medical Flu and COVID-19 Test | Microbiology | SIA | Yes | No | No |
| 2026-05-27 | K254075 | 510(k) | Fujifilm | Synapse Lung Nodule AI | Radiology | OEB | Yes | No | Yes |
| 2026-05-27 | K260680 | 510(k) | Philips | EPIQ Series Diagnostic Ultrasound Systems | Radiology | IYN | No | No | No |
| 2026-05-28 | K261405 | 510(k) | Imeka Solutions | ANDI 2.2 | Radiology | QIH | Yes | No | No |
| 2026-05-29 | K252859 | 510(k) | iRhythm Technologies | ZEUS Platform (FG0501US) | Cardiovascular | DQK | Yes | No | No |
| 2026-05-29 | K252947 | 510(k) | HistoSonics | HistoSonics® Planning Tool | Radiology | QTZ | Yes | No | No |
| 2026-05-29 | K260077 | 510(k) | OrthoGrid Systems | OrthoGrid Hip AI® 4.0 | Radiology | QIH | Yes | No | Yes |
| 2026-05-29 | K261369 | 510(k) | Cosmo AI | GI Genius™ Module 300 (GGM300-US) | Gastroenterology / Urology | QNP | No | No | No |
| 2026-06-02 | K252844 | 510(k) | TytoCare | Tyto Insights for Wheeze Detection (with PCCP) | Anesthesiology | PHZ | Yes | Yes | Yes |
| 2026-06-03 | K260780 | 510(k) | DESKi | HeartFocus | Radiology | QJU | Yes | Yes | Yes |
| 2026-06-04 | K253984 | 510(k) | Medtronic | Instrument Exit Point on Touch Surgery™ Aide | Surgery | SFE | Yes | No | No |
| 2026-06-04 | K260497 | 510(k) | Keya Medical | DEEPVESSEL Plaque | Radiology | QIH | Yes | No | Yes |
| 2026-06-04 | K261519 | 510(k) | Sonio | Sonio Suspect | Radiology | POK | Yes | No | No |
| 2026-06-05 | K254086 | 510(k) | Oculogica | EyeBOX SNAP | Neurology | QEA | No | No | Yes |
| 2026-06-08 | K260332 | 510(k) | Pathway Labs | EchoNext | Cardiovascular | QXO | Yes | No | Yes |
| 2026-06-09 | K260748 | 510(k) | E-Scopics | ES-Series | Radiology | IYN | Yes | No | Yes |
| 2026-06-10 | K253192 | 510(k) | Alpha Intelligence Manifolds | DeepXray Spina | Radiology | SAO | Yes | No | Yes |
| 2026-06-12 | K260406 | 510(k) | Brainomix | Brainomix 360 Hyperdensity | Radiology | QIH | Yes | Yes | No |
| 2026-06-15 | K253172 | 510(k) | Crescom | MediAI-OA | Radiology | QIH | Yes | No | No |
| 2026-06-15 | K253690 | 510(k) | Siemens Healthineers | LungMaps | Radiology | QIH | Yes | No | No |
| 2026-06-15 | K253825 | 510(k) | DeepHealth | Saige-Dx | Radiology | QDQ | Yes | No | Yes |
| 2026-06-15 | K260300 | 510(k) | Anumana | WatchMate Software | Radiology | QIH | Yes | No | Yes |
| 2026-06-16 | K260844 | 510(k) | Siemens Healthineers | ACUSON Sequoia Diagnostic Ultrasound System | Radiology | IYN | No | No | No |
| 2026-06-18 | K253050 | 510(k) | Informai | RadOncAI | Radiology | MUJ | Yes | No | Yes |
| 2026-06-18 | K260190 | 510(k) | ROPCA | ARTHUR | Radiology | IYO | No | No | Yes |
| 2026-06-18 | K260729 | 510(k) | GuideAI Health | Vascular Assist Occlusion Triage (VAOT) (v1.0) | Radiology | QAS | Yes | No | Yes |
| 2026-06-18 | K261713 | 510(k) | Fujifilm | Synapse PACS (7.6.0) | Radiology | QIH | Yes | No | No |
| 2026-06-22 | K253148 | 510(k) | Rhythm Express | RX-1 Sleep | Anesthesiology | MNR | Yes | No | No |
| 2026-06-24 | K253560 | 510(k) | GE HealthCare | Enhanced Boundary for PCCT | Radiology | JAK | No | No | No |
| 2026-06-24 | K260998 | 510(k) | Ceribell | Ceribell Neurology Review Software | Neurology | OMB | Yes | No | Yes |
| 2026-06-25 | K253786 | 510(k) | Siemens Healthineers | syngo.CT Coronary Cockpit | Radiology | QIH | Yes | Yes | Yes |
| 2026-06-26 | K260714 | 510(k) | Therapixel | MammoScreen® (5) | Radiology | QDQ | Yes | No | No |
| 2026-06-27 | K253390 | 510(k) | HoneyNaps | SOMNUM (SOMNUM) | Neurology | OLZ | Yes | No | Yes |
| 2026-06-29 | K253628 | 510(k) | Agada Medical | Auto-Seg (SO-0012), Spine Auto-Seg (SO-0012) | Radiology | QIH | Yes | No | No |
| 2026-06-29 | K253969 | 510(k) | ADAS 3D Medical | ADAS 3D | Radiology | QIH | Yes | No | No |
| 2026-06-29 | K260667 | 510(k) | Philips | Alturion Series Diagnostic Ultrasound System | Radiology | IYN | No | No | Yes |
