PET Image Reconstruction Patents: Who Leads, Trends 2026
- 22.5% of all filings sit with five companies, yet the ranked list runs to 100 names, pointing to a long tail below the leaders.
- Filings fell 34% from 2021 to 2024, the last year with a complete publication record, after peaking in 2019 at 884.
- G06T image data processing appears in 44.4% of records, while AI-model computing (G06N) sits at just 10.4%, well behind the imaging-hardware classes.
Filing growth compares 2021 (777 records) with 2024 (512) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 12,533 records in scope (CR5), not by the ranked leaders only.
What the PET reconstruction patent record shows
PET image reconstruction covers the algorithms and hardware pipelines that turn raw coincidence data into a diagnostic image: iterative reconstruction, point spread modelling, regularization strategies, and increasingly, deep learning denoising layered onto conventional reconstruction. The dataset in scope spans 12,533 published records filed between 2015 and mid-2026, drawn from a search string that pairs reconstruction methods with the specific technical levers used to control noise texture, quantitative accuracy and reconstruction time.
Filing activity concentrates around a handful of large medical imaging OEMs, but the presence of 100 ranked assignees, most with modest counts, indicates the field is not closed to new entrants. The IPC composition below shows where claim density sits today and where it thins out.
Filing trends and technology composition
The two views below cover the same 12,533 records: one across time, one across IPC subclass. Both use the full record set as the denominator, so class shares add up to more than 100% because a single record often carries several classes.
A peak in 2019, then a real pullback through 2024
Filings ran 789 in 2017, climbed to a peak of 884 in 2019, then declined to 512 by 2024 — a 34% drop from the 777 filed in 2021. Counts for 2025 and 2026 are still low because publication typically lags filing by around 18 months; they should not be read as a continuing collapse until later data fills in.
Image data processing dominates, AI-model classes still a minority
G06T (image data processing & generation) appears in 44.4% of records and A61B (diagnosis & surgery) in 35.4%, reflecting the field's roots in imaging hardware and clinical workflow. G06N (AI-model computing) sits at 10.4% and G06V (image/video recognition) at 5.8%, showing that deep-learning-labelled claims are still a minority layer on top of conventional reconstruction math.
Shares are the percentage of the 12,533 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on PET Image Reconstruction with Eureka
This page is one run against one query. Ask Eureka your own question about pet image reconstruction and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative and most-cited filings
Iterative image reconstruction with a sharpness driven regularization parameter
A method includes performing a first pass of an iterative image reconstruction in which an intermediate first spectral image and an intermediate second spectral image are generated using an iterative image reconstruction algorithm, start first spectral and second spectral images, and initial first spectral regularization and second spectral regularization parameters, updating at least one of the initial first spectral regularization or second spectral regularization parameters, thereby creating an updated first spectral regularization or second spectral regularization parameter, based at least on a sharpness of one of the intermediate first spectral or second spectral images, and performing a second pass reconstruction using the updated parameter.Filed by Philips, published 2017-06-29.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20100111370A1 | Method and apparatus for estimating body shape | 957 |
| 2 | US20180075581A1 | Super resolution using a generative adversarial network | 863 |
| 3 | US20030216630A1 | Conductivity reconstruction based on inverse finite element measurements in a tissue monitoring system | 644 |
| 4 | US20110044524A1 | Tool for accurate quantification in molecular MRI | 551 |
| 5 | US6026173A | Electromagnetic imaging and therapeutic (EMIT) systems | 497 |
| 6 | US20070083114A1 | Systems and methods for image resolution enhancement | 479 |
| 7 | US6490476B1 | Combined PET and X-ray CT tomograph and method for using same | 463 |
| 8 | US20040054248A1 | Radioactive emission detector equipped with a position tracking system | 370 |
| 9 | US6005916A | Apparatus and method for imaging with wavefields using inverse scattering techniques | 368 |
| 10 | US20180144243A1 | Hardware system design improvement using deep learning algorithms | 364 |
Citation counts favour older filings simply because they have had more time to accumulate them; treat this table as a map of influence, not of current commercial relevance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the numbers mean for filing strategy
Three patterns stand out once the raw counts are read against their denominators.
Leadership is real but not dominant
The top 5 assignees combined hold 22.5% of all 12,533 records in scope, and the top 10 hold 30.8%. That leaves close to 70% of filings spread across the remainder of the 100 ranked assignees, a pattern more consistent with an open competitive field than a locked oligopoly.
A genuine pullback, not a data artefact
Filings dropped from 777 in 2021 to 512 in 2024, the last year complete enough to trust. That decline sits well before the publication-lag window, so it reflects an actual slowdown in new filing activity rather than incomplete recent data.
Imaging math still outweighs AI framing
G06T (image data processing) covers 44.4% of records against G06N (AI-model computing) at 10.4%. Deep learning denoising is present in the corpus but is still filed mostly as an add-on to conventional reconstruction claims rather than as a standalone AI-first architecture.
US and EPO carry the bulk of filing activity
The United States receives 6,071 filings, ahead of the EPO at 2,012 and WIPO/PCT at 1,557. China trails at 554, suggesting most applicants are still filing this technology through US or European offices first rather than China-first.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to pet image reconstruction, with the prior art for and against each one.
Who is filing, and where the field is thinning out
The leader board is topped by large medical imaging OEMs, but momentum data suggests even those leaders have slowed their recent filing pace.
A single OEM well ahead of the field
The leading assignee holds 1,275 records, more than four times the fifth-place count of 302. That gap is the clearest sign of sustained, multi-generation investment in reconstruction IP rather than opportunistic filing.
A steep drop-off after the top handful
Fifth place holds 302 records and tenth place 166 — nearly half, in a span of only five ranking positions. Below tenth place, the ranking runs out to 100 assignees, many with far smaller counts, forming a long competitive tail.
Leaders are pulling back on new filings
Recent-year momentum for several top assignees is sharply negative: one large OEM shows -77% YoY, another -86%, and multiple leaders show zero filings in the latest year. Given publication lag, some of this reflects filings still working through the pipeline, but the direction across several leaders at once is consistent.
Collaboration is limited and mostly intra-group
Only 10 co-assignee pairs appear in the dataset, and the strongest pairs link related entities within the same corporate family rather than independent co-development. Cross-company joint filing is not a significant feature of this field yet.
| Assignee | Recent year | YoY |
|---|---|---|
| GE Precision Healthcare LLC | 3 | -77% |
| Shanghai United Imaging Healthcare Co., Ltd. | 2 | -33% |
| The Regents of the University of California | 1 | -86% |
| Koninklijke Philips N.V. | 0 | -100% |
| General Electric Company | 0 | — |
| Siemens Healthineers AG | 0 | — |
| Seno Medical Instruments, Inc. | 0 | — |
| Canon Medical Systems Corporation | 0 | -100% |
Where to take this next
The dataset points to specific follow-up questions rather than a single conclusion.
Check freedom-to-operate before filing in core reconstruction claims
With 22.5% of records held by five assignees and a further long tail beneath them, a targeted search against the leading assignees' active families is the first step before drafting new reconstruction claims.
Run a freedom-to-operate search in Eureka →Track the AI-denoising layer separately from core reconstruction math
G06N-tagged filings remain a minority relative to G06T, which suggests the AI-denoising layer is still being claimed as an add-on. Monitoring how that ratio shifts year over year flags when it becomes a standalone claim category.
Set up a monitoring watch in Eureka →Validate the 2021-2024 slowdown against your own competitor set
The field-wide 34% decline may not apply evenly across individual competitors; some leaders show near-zero recent filings while others may be filing steadily. Pulling assignee-level trend lines clarifies which pattern applies to any given competitor.
Compare assignee trend lines in Eureka →Common questions on PET image reconstruction patents
One assignee leads the ranked list with 1,275 records, well ahead of the fifth-placed assignee at 302 and the tenth at 166. That said, the top 5 assignees combined hold only 22.5% of the 12,533 records in scope, and the ranking extends to 100 assignees, so leadership is concentrated at the very top but the wider field remains open. Anyone assessing competitive position should look at both the leader gap and the size of the tail below it.
Filings peaked in 2019 at 884 and then declined, falling 34% from 777 in 2021 to 512 in 2024, the most recent year with a complete publication record. Counts for 2025 and 2026 look lower still, but that is expected because publication typically lags filing by about 18 months, so those years are not yet fully reported. Based on the complete years, the honest read is a genuine pullback from the 2019 peak rather than continued growth.
The AI-model computing class (G06N) appears in 10.4% of the 12,533 records in scope, compared with 44.4% for general image data processing (G06T). This indicates deep learning denoising and related AI techniques are present but still a minority layer, usually claimed alongside conventional iterative reconstruction methods rather than as a fully standalone architecture. Filers should not assume AI framing dominates claim language in this space yet.
Branches such as deep learning denoising layered onto iterative reconstruction, point spread function modelling specific to time-of-flight PET, reconstruction-time optimisation for on-scanner or edge compute, and cross-modality regularization combining PET with MR or CT priors all show thinner claim density relative to the core G06T and A61B classes. These are not guaranteed clear, but they warrant a focused freedom-to-operate check rather than being assumed crowded by default. A first claim in these areas should tie the specific technical mechanism (e.g., a named regularization prior) to a measurable outcome like reconstruction time or noise texture.
The United States receives the largest share, with 6,071 filings, followed by the European Patent Office at 2,012 and the WIPO/PCT route at 1,557. China trails at 554 and Germany at 461, indicating that most applicants in this field currently prioritise US or European filing routes ahead of a China-first strategy. This pattern is worth revisiting periodically since receiving-office priorities can shift as regional imaging markets grow.
Research PET Image Reconstruction in depth with Eureka
Go past this page: query the whole pet image reconstruction corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.