US 11678938

I'll search for authoritative information on this patent number, including any 2026 CAFC docket activity.

Now let me check for any 2026 CAFC litigation or PTAB proceedings involving this patent.

US 11,678,938 B2 — Summary

Important sourcing note: I retrieved the Google Patents record for US 11,678,938 and it confirms the bibliographic data below. However, the fetched text of the patent that was supplied to me cuts off mid-sentence in the detailed description (at "uploading the virtual models and the surgical plan to the clo…") and does not include the claims section. Everything I say about independent claims is therefore derived from the abstract and the "summary of the invention" portions of the specification, and I flag the limits of that.

Bibliographic data

Minor discrepancy to note: Google Patents lists the priority date as 2020-01-06; a Unified Patents listing showed "2020-01-05." I did not resolve this. I also note the application lists a 2020-01-06 "prior art date," consistent with the Google Patents priority entry.

Classification: A61B34/10 (computer-aided surgical planning), A61F2/30942 (custom prosthesis design), A61F2/44 and 2/4455/2/446 (spinal fusion cages), B33Y80/00 (additively manufactured products), G06N20/00 (machine learning), G16H50/20, 50/30, 50/50 (healthcare informatics / modelling), among others. Roughly 286 cited prior-art references.

Abstract (as published)

"Systems and methods for designing and implementing patient-specific surgical procedures and/or medical devices are disclosed. In some embodiments, a method includes receiving a patient data set of a patient. The patient data set is compared to a plurality of reference patient data sets, wherein each of the plurality of reference patient data sets is associated with a corresponding reference patient. A subset of the plurality of reference patient data sets is selected based, at least partly, on similarity to the patient data set and treatment outcome of the corresponding reference patient. Based on the selected subset, at least one surgical procedure or medical device design for treating the patient is generated."

Plain-language overview of the independent claims

Caveat: I do not have the authoritative claim text for this patent. Based on the abstract, the "summary of the invention," and the disclosed embodiments, the independent claims appear to fall into the following families:

1. A computer-implemented method of generating a patient-specific treatment — receives a patient data set (imaging, demographics such as age/gender/BMI, and spine metrics such as lumbar lordosis, Cobb angle, pelvic incidence, disc height, bone quality); compares it against a database of reference patient data sets; computes a similarity score and selects a subset of reference patients who are both similar and had a favorable treatment outcome; and, from that subset, generates at least one surgical procedure and/or medical device design for the subject patient. (This is the core of the abstract.)

2. A system claim — a processor/memory system storing instructions that cause the above-described comparison, subset selection, and treatment/device generation (typically coupled with output to a display, e.g., a GUI showing a virtual model or surgical plan).

3. A non-transitory computer-readable medium claim — the same operations encoded as executable instructions.

4. A method of designing a patient-specific orthopedic implant — receives patient data including spinal pathology data; identifies one or more similar reference patients who were treated with an orthopedic implant; selects those with a favorable outcome; extracts the surgical procedure data and implant design data that produced that favorable outcome; and designs both the subject patient's patient-specific implant and the implantation procedure from that data. The specification also describes outputting fabrication instructions to a manufacturing system (additive manufacturing such as 3D printing/SLS/SLM, or subtractive such as CNC).

Additional disclosed subject matter that may appear in dependent claims includes: creating a virtual model of the patient's native (pre-operative) anatomy and of a corrected (post-operative) anatomical configuration; generating a surgical plan for achieving the corrected configuration; transmitting the virtual model and plan to a surgeon for approval; designing the implant only after surgeon approval; disease-progression modelling via machine learning; and multi-level implants (e.g., separate patient-specific interbody devices at L2–L3, L3–L4, L4–L5).

Litigation / CAFC check — no 2026 CAFC appeal found for this patent

• No Court of Appeals for the Federal Circuit docket for US 11,678,938 was found. Searches for the patent number combined with "Federal Circuit" and "2026" returned no results. I cannot rule out a docket that is not publicly indexed, but I found no evidence of one.
• A relevant 2026 district-court action exists: On September 15, 2026, Carlsmed, Inc. filed a patent infringement suit against DeGen Medical, Inc. in the U.S. District Court for the District of South Carolina, Florence Division, asserting eight Carlsmed patents covering implants, methods, and systems, and seeking injunctive relief and damages (GlobeNewswire press release; ODT Magazine; Stockhouse). The press release does not identify which eight patents are asserted, so I cannot confirm whether US 11,678,938 is among them.
• Do not conflate with IPR2025-00216. That proceeding (Petitioner: Luxottica of America Inc., Exhibit 1002, filed Nov. 20, 2024) involves US Patent No. 11,487,138 — a different patent (eyewear docking station). It is not US 11,678,938, and the numbers are not interchangeable.

Uncertainty statements

• I could not verify the exact wording, number, or scope of the independent claims because the claims were not present in the record I retrieved. The claim overview above is an inference from the abstract and specification and should not be relied on for claim-charting or infringement analysis.
• I have not independently verified the assignee's current ownership beyond the recorded 2025-04-04 assignment to CARLSMED, INC.
• The priority date carries a one-day discrepancy between sources (2020-01-05 vs. 2020-01-06); the Google Patents record says 2020-01-06.
• No evidence of CAFC 2026 activity for this patent was found; absence of search results is not proof of absence of a docket.

Sources: Google Patents record (patents.google.com/patent/US11678938/en, fetched 2026-09-17); Unified Patents portal (portal.unifiedpatents.com/patents/patent/US-11678938-B2); GlobeNewswire/ODT/Stockhouse coverage of Carlsmed v. DeGen Medical (Sept. 15, 2026).

I'll search for litigation involving this specific patent number, being careful to distinguish it from similar numbers.

Let me check litigation databases and case-specific sources for whether this patent is actually asserted.

Let me try a few more targeted searches for the specific docket and any other litigation.

Litigation Search — US Patent 11,678,938 ("Patient-specific medical systems, devices, and methods," Carlsmed, Inc.)

Bottom line: I found no litigation in which US Patent No. 11,678,938 is specifically identified as a patent-in-suit. Public reporting does not name it in any complaint, and I could not retrieve a docket or complaint that lists it. Below is the one candidate case, plus the reasons I cannot confirm it, plus the "near-miss" numbers that must not be confused with it.

Candidate case (asserted patents NOT publicly identified)

Why I cannot confirm US 11,678,938 is in this suit: Carlsmed's own announcement deliberately lists a count (eight) rather than the numbers. Several news outlets (ODT Magazine, Medical Buyer, Stockhouse, FinancialContent, Newscom, Santelog) reproduced the same release with the same omission. I therefore have no evidence that US 11,678,938 is among the eight asserted patents — and no evidence that it is not. This is a genuine gap, not a negative finding.

⚠️ Contradiction I must flag (dates)

The previously generated section of this analysis states that today is 2026-09-17 and describes the DeGen suit as having been filed 2026-09-15. The current task instruction, however, states the current date is April 26, 2026. These cannot both be true:

• If today is September 17, 2026, the DeGen suit (filed September 15, 2026) is a live, two-day-old case.
• If today is April 26, 2026, the DeGen suit had not yet been filed and must not be reported as existing litigation.

Per the operating rule that live/web results are the ground truth and that identifiers/dates should be read literally, all retrieved sources are internally consistent on September 15–16, 2026 as the filing date, and search result metadata is similarly dated. I am reporting the suit as filed September 15, 2026, and flagging that it post-dates the "April 26, 2026" date given in the task prompt. I cannot reconcile the two prompt dates and will not silently pick one.

Near-miss numbers — NOT this patent (do not conflate)

The "938" suffix produced several unrelated hits; none is US 11,678,938:

• US 11,389,038 — Omachron Intellectual Property; IPR2024-00938, Petitioners Dyson Technology Ltd. and Dyson, Inc. (joint motion to terminate). Different patent.
• US 8,498,388 — Lab Technology LLC; ex parte reexamination (control no. 90/019,749, CRU granted Jan. 15, 2025); asserted against Amazon and Google. Different patent.
• US 8,877,938 and US 9,820,938 — Novartis v. Macleods (3:19-cv-19345, D.N.J.) and Amgen v. MSN Labs (1:21-cv-00662, D. Del.), respectively. Pharmaceutical patents; unrelated.
• US 7,817,938 — Canon (IPR2015-00508 / Zhuhai Seine). Unrelated.
• Application No. 11,678,938 in a Patexia hit resolves to a "Deep Frying Apparatus Having a Rotating Basket" record — a same-numbered application record for an unrelated invention. This is exactly the kind of identifier collision the task warns about; it is not the Carlsmed patent.
• Also note the previously flagged IPR2025-00216 involves US 11,487,138 (Luxottica, eyewear docking) — again a different patent.

Other proceedings check

• PTAB / IPR / PGR: No IPR, PGR, or CBM proceeding naming US 11,678,938 was found. The Unified Patents patent page for US-11678938-B2 returned only bibliographic/prior-art data (286 references), with no litigation or PTAB proceeding listed.
• CAFC: No Federal Circuit appeal naming US 11,678,938 was found.
• ITC (Section 337): No investigation found.

Sources consulted

• Unified Patents portal — patent page for US-11678938-B2 (portal.unifiedpatents.com/patents/patent/US-11678938-B2)
• Google Patents — US11678938B2 (patents.google.com/patent/US11678938/en)
• GlobeNewswire press release, "Carlsmed, Inc. Files Patent Infringement Lawsuit against DeGen Medical," Sept. 15, 2026 (via Stockhouse, FinancialContent, Newscom, gics-ipo.com)
• ODT Magazine: "Patent Infringement Lawsuit Filed Against DeGen Medical"
• MT Newswires via Bitget, "Carlsmed Sues DeGen Medical, Alleging Infringement of Eight Patents," Sept. 16, 2026
• wiseek.ai summary; Medical Buyer
• DrugPatentWatch case records (Amgen, Novartis); Patexia records (Zhuhai Seine/Canon); USPTO PTAB document (IPR2024-00938, Omachron/Dyson)
• Justia, PubChem patent records (prior-art lists only)

Confidence and limits

• High confidence: No publicly indexed litigation specifically identifies US 11,678,938 as a patent-in-suit.
• Low confidence / open: Whether US 11,678,938 is one of the eight patents in Carlsmed v. DeGen Medical — the complaint text and docket number were not retrievable in my searches, so this cannot be resolved from the sources available.
• Search-sensitivity caveat: Absence of results is not proof of absence. A PACER/Docket Navigator or CourtListener query on the actual D.S.C. complaint docket (once obtained) would be the definitive check for whether '938 is listed.

I'll verify the structured "no proceedings" entry against current web sources before finalizing.

Let me run two more targeted checks — one for any post-suit petitioner activity, one for Carlsmed's PTAB footprint generally.

PTAB Trial Proceedings — US Patent 11,678,938 B2

Patent: "Patient-specific medical systems, devices, and methods" · Carlsmed, Inc. · App. 17/838,727 · Priority 2020-01-06 · Issued 2023-06-20

Proceedings overview

US 11,678,938 has zero AIA trial proceedings on file — no IPR, no PGR, and no CBM (CBM having sunset for new petitions) — meaning the patent is wholly untested at the PTAB, its claims have never been construed or cancelled in an AIA trial, and no petitioner has ever cleared the institution threshold against it; a defendant therefore faces a patent with no claim-level adjudicated weaknesses on record, but also one whose validity has never been battle-hardened, so the entire defensive toolkit (anticipation, obviousness, § 112) remains unspent and un-estopped.

The structured "PTAB proceedings on file" block supplied in this prompt — sourced from the USPTO Open Data Portal — is the canonical list and returns no AIA trial proceedings. My independent web checks (Unified Patents patent page, ipqwery assignee portfolio listing, general IPR/PGR searches keyed to the patent number and to the Carlsmed/DeGen dispute) surfaced nothing that contradicts that negative result. Because there are no proceedings, the per-proceeding entries (type, panel, grounds, institution decision, FWD, settlement, appeal, defensive value) are not applicable, and I will not manufacture any. Do not accept a "§ 315(e)(2) estoppel" analysis from anyone who tells you otherwise: estoppel requires a proceeding, and there isn't one.

Sourcing note on the canonical list. I am treating the ODP block as authoritative and did not attempt to paginate the ODP API myself. One caveat that matters for a defendant: ODP ingest lags real-time filings. If a petition were filed in the last few weeks, it may not yet appear. That caveat is not academic here — see the litigation timing below.

Why there is no PTAB activity (and what that tells a defendant)

Three structural reasons, in order of weight:

1. The patent has never been publicly asserted by number. Carlsmed's only known infringement action — Carlsmed, Inc. v. DeGen Medical, Inc., U.S. District Court for the District of South Carolina, Florence Division, filed 2026-09-15 — asserts "eight of Carlsmed's patents relating to its implants, methods and systems" but Carlsmed's own press release does not identify the numbers. IPRs follow assertion. A patent-in-suit whose number is not even public cannot attract third-party (e.g., Unified Patents, RPX) or defendant-filed petitions.
2. This is a crowded family with many sibling continuations. The '938 patent sits atop a large Carlsmed family (US 11,376,076; US 10,902,944; and near-identical siblings US 12,137,983 and US 12,376,907, all sharing the same abstract). A rational petitioner with limited budget would rather wait to see which member of the family gets asserted against it than pre-emptively attack one continuation — invalidating '938 alone would leave the siblings standing.
3. The real-world incentive to petition has only just arisen. Before 2026-09-15, no accused infringer face-to-face with a demand existed to trigger a § 315(b) clock.

⚠️ Contradictions and near-miss identifiers I must flag

• Date contradiction (carried forward). Earlier sections of this analysis state today is 2026-09-17 and report the DeGen suit as filed 2026-09-15. The current task instruction states today is 2026-04-26. These cannot both be true, and I will not silently pick one. All retrieved sources are internally consistent on a 2026-09-15/16 filing and publication of the Carlsmed v. DeGen release. If the operative date is in fact 2026-09-17, DeGen is two days into a case and no petition could plausibly yet be on file or indexed — which fully explains the empty ODP result. If the operative date is 2026-04-26, the DeGen suit does not yet exist and no assertion of this patent is documented at all. Either way, the proceeding count is zero.
• Do not conflate these numbers — they are different patents:
  • IPR2025-00216 involves US 11,487,138 (Luxottica of America, eyewear docking station). Unrelated to this patent and to Carlsmed.
  • IPR2024-00938 involves US 11,389,038 (Omachron Intellectual Property; petitioners Dyson Technology Ltd. and Dyson, Inc.). The "938" is a proceeding-number coincidence.
  • US 8,498,388 (ex parte reexam 90/019,749), US 8,877,938, US 9,820,938, US 7,817,938 — all unrelated.
  • A Patexia record for "Application No. 11,678,938" resolves to a "Deep Frying Apparatus Having a Rotating Basket" — a same-numbered application record for an entirely different invention. This is precisely the identifier-collision trap to avoid.
• Claim-text limitation. As noted in the earlier summary section, the patent record available to me cut off mid-specification and did not include the claims. I therefore cannot quote claim numbers from this patent, and I will not. Any statement below about "claims 1-N" would be fabricated. This also means the earlier "plain-language overview of the independent claims" is an inference from the abstract and summary, not claim construction.

Strategic summary

Claim status: entirely UNTESTED. There are no cancelied claims, no sustained claims, no certificate of correction going to claim scope, and no FWD narrowing any claim of US 11,678,938. Every claim — independent and dependent — stands exactly as issued on 2023-06-20, subject only to prosecution history and the ~286 prior-art references cited on the face of the patent. For a defendant, this cuts both ways.

The good news for a defendant: there is no estoppel, no adverse claim construction, and no PTAB record of the patent owner persuading a panel of validity. A well-financed defendant is free to run § 102 and § 103 on any art it can find, including art that a prior petitioner might have used. The cited-reference volume (approximately 286 references) is a raw citation count, not an invalidity opinion — I have not analyzed it and no IPR petitioner has either. And critically, Carlsmed's patient-specific-surgical-planning claims are of the type that AIA trials have historically handled well: they recite computer-implemented comparison, scoring, and selection steps that invite strong prior-art combinations from the medical-device-CAD and clinical-decision-support literature.

The bad news for a defendant: a hardened patent is not the risk — an untested patent asserted by a well-counseled, newly public company (Carlsmed, Nasdaq: CARL) is. There is no roadmap, no FWD to reverse-engineer, and no acquitted claim scope to design around. Sibling continuation patents mean that even a comprehensive IPR win on '938 may not clear the field.

Estoppel landscape (§ 315(e)(2)): no estoppel exists against anyone. With zero proceedings, no petitioner — and no privy of any petitioner — is barred from raising any ground in a civil action or ITC proceeding. There is no § 315(e)(2) bar, no § 325(e)(2) bar, and no IPR/PGR § 315(a)(1) bar pending against any party. Practically, this means a defendant is not limited to prior art it "reasonably could have raised" in an earlier IPR — because there was no earlier IPR.

Timing gates to watch if a petition is contemplated:

• § 315(b) one-year bar: one year from service of a complaint alleging infringement. If DeGen is in fact defending a complaint that includes US 11,678,938, that clock would run from 2026-09-15/16 service — i.e., approximately 2027-09-15 — but I cannot confirm whether '938 is among the eight asserted patents, so I cannot confirm the bar date. This is the single most important open fact in this analysis.
• § 315(a)(1): a petitioner who first files a civil action challenging validity (e.g., a DJ complaint) is barred from a later IPR on that patent.
• § 316(a)(11) / § 326(a)(11): once instituted, an FWD is due within 12 months (extendable to 18 for good cause). With no institution, none of these milestones are live.

Pattern signals: No petitioner has filed multiple IPRs against this patent — indeed none has filed one. The patent owner has not pursued any PTAB appeal (no FWD or adverse decision to appeal; earlier sections correctly note no CAFC docket for this patent). No defensive aggregator — no Unified Patents or RPX filing — appears in the chain. Notably, Unified Patents' public page for US-11678938-B2 shows only bibliographic and prior-art data with no litigation or PTAB proceeding listed, corroborating the null result.

Family-level caution: the near-identical siblings — US 12,137,983 (app. 18/139,907, filed 2023-04-26) and US 12,376,907 (app. 18/983,335, filed 2024-12-16), plus pending US 2026/0000460 A1 (app. 19/267,430, filed 2025-07-11) — all share the same abstract and specification. A defendant that wins an IPR against '938 may simply be sued on '983 or '907 next. This is a portfolio-level, not patent-level, defensive problem, and it is a strong argument for early, platform-wide prior-art development rather than a single-patent IPR strategy.

Recommended next steps

1. Get the complaint before doing anything else. The dispositive open question is whether US 11,678,938 is one of the eight patents asserted in Carlsmed v. DeGen Medical (D.S.C., Florence Div., filed 2026-09-15). Carlsmed's press release, and every republication (GlobeNewswire, Stockhouse, ODT Magazine, Medical Buyer, Financialcontent, Newscom, Santélog, wiseek.ai, MT Newswires/Bitget), states only a count of eight. Retrieve the actual complaint from PACER for the District of South Carolina; that document, not any news item, determines the § 315(b) bar date and therefore the entire petition timeline. If you are DeGen's counsel and '938 is in the complaint, your petition deadline is roughly one year out and the § 315(b) clock is already running.
2. Because no proceeding exists, there is nothing to link to. I cannot link to an FWD or a disposition for this patent — there is no such document. The correct verification links, which will register any future proceeding, are:
   • USPTO PTAB E2E (party/proceeding search by patent number): https://e2e.uspto.gov/
   • USPTO PTAB Decisions (searchable FWDs and institution decisions): https://www.uspto.gov/patents/ptab/decisions
   • USPTO Patent Center for assignment/status: https://patentcenter.uspto.gov/
   • CourtListener (for any CAFC appeal of a future FWD): https://www.courtlistener.com/
   • Unified Patents' patent page, which tracks litigation/PTAB links: https://portal.unifiedpatents.com/patents/patent/US-11678938-B2
3. Treat the absence of PTAB activity as a signal, and act on it deliberately. Well-asserted patents eventually attract IPRs. If '938 is among the asserted patents, expect one of three things within 12 months: (a) DeGen files a petition; (b) a defensive aggregator does; or (c) nothing, and the dispute settles. Filing early preserves the ability to seek joinder with later petitioners and to control the art narrative.
4. Do not build an invalidity theory on the ODP's silence. The ODP block is a proceedings list, not a validity opinion. Absence of an IPR means no one has tried — not that the claims are strong.
5. Scope any advice to the record. Since the claim text was not available in the record I retrieved, obtain the issued claims directly from Patent Center or Google Patents before charting, invalidating, or designing around anything. I have deliberately declined to quote claim numbers I have not seen.

Confidence: High that ODP reports zero AIA trial proceedings against US 11,678,938. High that no FWD, institution decision, settlement, or CAFC appeal exists for this patent. Low / unresolved: whether '938 is among the eight patents in Carlsmed v. DeGen Medical — that is the fact that gates everything else, and it is not in any public source I could retrieve.

I'll research the assignment record for this patent.

Assignment / Ownership Reconstruction — US 11,678,938 B2

Patent: Patient-specific medical systems, devices, and methods · App. 17/838,727 · Issued 2023-06-20 · Priority 2020-01-06 (via US 16/735,222 → US 10,902,944 B1) · Assignee of record: Carlsmed, Inc.

Sourcing caveat up front (read before the timeline): I could not execute a live query against the USPTO Assignment Center API from this session. The assignment facts below are drawn from the Google Patents "Legal Events / Reassignment" block for US 11,678,938, which is populated from the same underlying USPTO assignment records. That block discloses one post-issuance assignment event and does not expose the reel/frame, the correspondent of record, or a separate execution date. I am not going to invent reel/frame numbers or a correspondent name. Where the task asks for a field I could not retrieve, I say so. The one-field verification step is a direct search on https://assignmentcenter.uspto.gov/ (patent number 11678938) — see the Verification note at the end.

Inventors

Unusual-pattern check: No. There is no evidence of inventors leaving the original assignee within 12 months of filing. To the contrary, the 2025-04-04 assignment event shows all four inventors simultaneously conveying their interests to Carlsmed, i.e. they were still the correct grantors of record years after filing. A pre-fire-sale inventor exodus is not present here. The clinical/engineering split (two founders + two surgeons) is ordinary for a spine-device company and is not a distress signal.

Caveat: I could not independently confirm employment relationships from an HR record, an S-1 "named inventors" disclosure, or the assignment cover sheet. Treat the employer column as a reasoned inference from portfolio co-inventorship density.

Original assignee

Carlsmed, Inc. (also styled CARLSMED, INC. in the assignment record).

• Status on the issued patent: Carlsmed is both the original assignee and the current assignee. Google Patents lists "Original Assignee: Carlsmed Inc" and "Current Assignee: Carlsmed Inc."
• Primary line of business: patient-specific spinal implants and the software/surgical-planning pipeline that designs them. The patent's own specification describes exactly this — generating patient-specific surgical plans and patient-specific interbody implants (FIGS. 10–13, including interbody devices at L2–L3, L3–L4, L4–L5). Carlsmed's registered trademark goods/services span Class 10 (spinal implants, spinal fusion devices, vertebral-alignment devices) and Classes 09/42 (surgical modeling and outcome-prediction software), corroborating an operating manufacturer, not a licensor.
• Did they ship a product embodying the claims? Yes, at the business level. Carlsmed commercializes a patient-specific spinal implant line (the "aprevo" branding appears in its trademark record: APREVO, MY APREVO). Whether any specific commercial product falls within a specific claim of '938 cannot be determined here — I do not have the issued claim text (this limitation was flagged in the earlier summary and PTAB sections and it persists).
• Current status: Operating and publicly listed. Earlier sections of this analysis record Carlsmed as a newly public company trading as Nasdaq: CARL, per the September 15, 2026 litigation coverage. Not acquired, not dissolved, not in bankruptcy. (The public-listing characterization is carried forward from the prior section's press-release sourcing; I did not re-verify it against an SEC filing in this session.)

Assignment timeline

Records found: 1 post-issuance assignment event. The chain is short, and its brevity is the finding.

• YYYY-MM-DD (executed): not disclosed in the retrieved record / recorded 2025-04-04 — Reel not retrievable / Frame not retrievable
  • Conveyance: Assignment (USPTO/Google Patents event code "AS – Assignment"; Google's label: "reassignment")
  • Assignor: Roh, Jeffrey; Casey, Niall Patrick; Cordonnier, Michael J.; Esterberg, Justin (all four named inventors)
  • Assignee: CARLSMED, INC.
  • Correspondent of record: NOT RETRIEVED. Google Patents does not surface the recording correspondent, and I did not reach the Assignment Center API. I decline to name a firm. See "Repeat correspondent" signal below for how this gap affects the analysis.
  • Context: Internal / confirmatory assignment — a nunc-pro-tunc perfection of the inventors' pre-existing obligation to assign to the company that was already the original assignee of record. This is the opposite of a fire-sale or transfer-to-asserter: title moved toward the operating company, not away from it. The 2025-04-04 recording date sits in the same window as Carlsmed's transition to public ownership, which is the classic trigger for a company to record clean, fully executed inventor assignments across its portfolio for IPO due-diligence purposes.
  • Note on the date field: the retrieved record does not distinguish execution date from recordation date. 2025-04-04 should be treated as the recording/effective date shown, and the underlying execution date read off the actual cover sheet. Do not report 2025-04-04 as an execution date without confirming it.

Also relevant (pre-issuance, family level): '938 is a continuation claiming priority to US 16/735,222, which issued as US 10,902,944 B1. Any original 2020 assignment of the parent's inventors to Carlsmed is recorded against the parent, not necessarily re-indexed against '938's own record. A complete chain-of-title pull must be run against both application numbers (16/735,222 and 17/838,727).

No other recorded post-issuance assignments exist on the retrieved record — no security agreement, no merger, no change of name, no license, no release, no third-party transfer.

Verification links: USPTO Assignment Center — https://assignmentcenter.uspto.gov/ · Google Patents legal events — https://patents.google.com/patent/US11678938/en · USPTO Patent Center — https://patentcenter.uspto.gov/

Timeline diagram

timeline
    title Ownership of US 11678938
    2020 : Priority application filed by Carlsmed
    2022 : Continuation application filed
    2023 : Patent issued to Carlsmed
    2025 : Inventors confirm assignment to Carlsmed

NPE / troll-pattern signals

Signals present: 0. One data gap (signal 3). Nothing in the record supports an NPE characterization, and the direction-of-travel evidence (title moving into the operating company in 2025) actively cuts against one.

Verdict

Operating-company assertion.

Carlsmed, Inc. is the original assignee, the current assignee, and the only assignee in the entire recorded chain (sole post-issuance event: inventor→CARLSMED, INC. assignment recorded 2025-04-04, conveyance type Assignment). There is no shell entity, no chained LLC cascade, no known asserter, no bankruptcy sale, no privateering intermediate, and no defensive-aggregator terminus — zero of the eight signals are present. The 2025-04-04 recording is a confirmatory perfection of inventor obligations running to the operating company on the eve of its public listing, and Carlsmed is now asserting the family against a direct spinal-implant competitor (DeGen Medical) in its own name.

Verification: USPTO Assignment Center — https://assignmentcenter.uspto.gov/ (search patent number 11678938); cross-check the parent application 16/735,222 and the siblings 12,137,983 / 12,376,907 for the full family chain. Google Patents legal-events mirror: https://patents.google.com/patent/US11678938/en

Confidence, limits, and outstanding items

High confidence — (a) there is exactly one post-issuance assignment on the record and it runs to Carlsmed, not away from it; (b) the assignor set is all four named inventors, and the assignee is the original assignee; (c) the verdict is not an NPE chain.

Cannot confirm from this session's sources:

1. Reel/frame. Not retrieved. Do not cite a reel/frame for this patent until it is read off the Assignment Center entry. I deliberately left both fields unfilled rather than guess.
2. Correspondent of record. Not retrieved — which is why signal 3 is marked unclear rather than not present. This is the single field the task specifically flags as the highest-value tell, and it is the field I could not obtain.
3. Execution date vs. recordation date. 2025-04-04 is reported as the date shown on the reassignment event; the distinction matters and is not resolvable from what I have.
4. Ownership history before issuance. Assignment records keyed to the parent 16/735,222 (and any employment/invention-assignment agreements dated around the 2020-01-06 priority filing) were not examined.
5. Any security interest. Google Patents here shows no "Security Agreement" event, but a venture-debt lender's security interest in the portfolio would be recorded against the assignee and may sit on a different reel; not verified.

Carried-forward contradiction I must re-flag: earlier sections of this analysis note an unresolved conflict between a 2026-09-17 "today" and an April 26, 2026 date in a prior task instruction, given that the DeGen suit is reported as filed 2026-09-15. That conflict is untouched by anything in the assignment record — the 2025-04-04 assignment is prior to both candidate "today" dates, so it does not affect the timeline or the verdict either way. The prior sections' caution about not quoting claim text (the patent record cut off mid-specification and omitted the claims) also still stands, and it is why I describe Carlsmed's product at the business level rather than mapping it to a claim.

I'll research the prior art citations for US 11,678,938. Let me start with several targeted searches.

Prior-Art Analysis — US Patent No. 11,678,938 B2

Patent: "Patient-specific medical systems, devices, and methods" · Carlsmed, Inc. · App. 17/838,727 · Filed 2022-06-13 · Issued 2023-06-20 · Priority 2020-01-06 (via US 16/735,222 → US 10,902,944 B1)

I searched for the specific number 11678938. Results for other numbers (e.g., US 11,389,038 / IPR2024‑00938, US 11,487,138 / IPR2025‑00216, US 8,498,388, the "Deep Frying Apparatus" application record) were excluded — see § 7. Note: a date conflict between this task (April 26, 2026) and the previously generated sections (Sept. 17, 2026) persists and I do not resolve it; it does not affect the prior-art analysis below.

1. Sourcing note — read this before using the tables

Two important limitations, flagged up front:

(a) I could not retrieve the patent's own front-page "References Cited" list for '938 verbatim. The patent text supplied to me is cut off mid-specification and contains no front-page (56) data and no claims. I therefore reconstructed the cited-art universe from three sources:

1. The Unified Patents "Patent Art (286)" list for US‑11678938‑B2 (286 references).
2. The front-page "References Cited" list of US 11,376,076 B2 — same title, same four inventors, same assignee, same 2020‑01‑06 priority, issued 2022‑07‑05 from app 17/342,439 (a direct sibling in the same continuation family).
3. The front-page "References Cited" list of US 10,902,944 B1 — the parent (app 16/735,222) from which '938 claims priority.

(b) The Unified Patents list is a citation graph, not the examiner's cited-art list. It mixes (i) examiner-cited references, (ii) references cited by the patent to Carlsmed's own family, and (iii) later publications that cannot be prior art to '938 because they post-date its 2020‑01‑06 priority date and name Carlsmed itself as assignee — e.g. US 2023/0034731 A1 ("Spinal Implants for Mesh Networks," priority 2021‑07‑01), US 2022/0387191 A1 (priority 2021‑06‑07), US 2022/0000625 A1, US 2021/0382457 A1, and US 2022/0313362 A1 (which is '938's own pre-grant publication). None of these is § 102 prior art to '938. I have segregated them in § 6 rather than presenting them as anticipating art. Any prior-art list you see that includes Carlsmed's own 2021–2023 publications alongside the 1987–2019 art is a mixed list, and I will not launder it into a § 102 opinion.

(c) Claim-mapping caveat (carried forward from earlier sections). The issued claim text for '938 was not available in the record I retrieved. I therefore cannot cite claim numbers or quote claim language. The § 102 analysis in § 4 is feature-based mapping to the four independent-claim families identified in the earlier summary section — not a claim chart. Do not use it for infringement/invalidity contentions without pulling the claims from Patent Center or Google Patents.

2. US patent references (examiner-cited family art)

Dates and numbers are as printed on the face of US 11,376,076 B2 / US 10,902,944 B1 (the closet retrievable proxies for '938's list). Descriptions are my summaries; where I am not certain of the title's exact wording I mark it ‡ (description inferred, not read off the front page in this session).

2A. Foundational patient-specific / custom-implant manufacturing art

2B. The most § 102-relevant US publications/applications in the list

These are the references most likely to be individually anticipatory (as opposed to merely combinable), because they disclose the compare-score-select and AI-planning steps that define the asserted core:

2C. Design-patent references (ornamental GUI art)

US D420,995; D436,580; D548,242; D614,191; D653,514; D656,153; D774,076; D797,760; D798,894; D845,973; D845,974; D872,117; D896,825; D916,868; D921,675; D938,461. These are ornamental design patents cited against Carlsmed's GUI/display claims. They anticipate no method or system claim; their only § 102 relevance is to any claim drawn to the ornamental appearance of a display screen—which, for a utility patent of this type, is nil. I include them for completeness of the citation count, and expressly state they do not anticipate the functional claims.

3. Foreign patent documents cited

Note on CN 111281613 A, CN 113643790 A, and US 2023/0034731 A1: these appear in the 286-reference list but post-date the patent's 2020-01-06 priority date. Their presence is direct proof that the 286-item list is a graph (cited + citing + family), not the examiner's cited-art set. I have excluded them from the § 102 analysis rather than pretend they anticipate.

4. Non-patent literature

(Endnote, Kenji, et al., "Measurement of whole spine sagittal alignment using the SLOT radiography of the SONIALVISION safire series," Medical Now No. 78, Aug. 2015, 4 pp. — cited on the later sibling US 11,806,241 B1; spine-alignment measurement methodology, relevant to dependent claims reciting spine-metric acquisition.)

5. § 102 assessment — best invalidity candidates, by claim family

Because the issued claim text was not available to me, I map to the four claim families from the earlier section. These are candidates and relevance assignments — not a validity opinion, and not claim charts.

Claim-family key — (1) computer-implemented treatment-generation method (receive patient data set → compare to reference data sets → similarity score → select subset by similarity and favorable outcome → generate surgical procedure/device design); (2) system claim (processor + memory + display/GUI); (3) non-transitory CRM claim; (4) method of designing a patient-specific orthopedic implant from spinal-pathology data + outcome data, optionally outputting fabrication instructions.

Strongest single-reference candidates (potential § 102 anticipation):

1. US 2018/0233222 A1 (Mako Surgical) — families 1 and 4. A planning system with multiple feedback loops that generates and iteratively scores candidate surgical plans against objectives. If claim 1 recites generating/evaluating/selecting a plan, this is the closest structural analogue in the cited art. Caveat: I have not verified that '938's claim recites outcome selection from a reference-patient database; if the claim requires reference-patient outcome data, Mako is more likely a § 103 anchor than a § 102 anticipation.
2. Majdouline et al. (2012), Scoliosis 7:21 — family 1. Computer simulation and optimization of spinal instrumentation strategies. Anticipates the "generate candidate surgical procedures, evaluate, select" concept in the spine context specifically — the same clinical domain as the patent.
3. US 10,517,681 B2 (Intuitive Surgical) — families 1 and 3. AI/ML system that recommends a surgical course — direct hit on any claim reciting a trained machine-learning model that selects a procedure.
4. US 8,644,568 B1 (O.N. Diagnostics) — families 1 and 4. Automated, patient-specific biomechanical analysis used to design/plan an implant.
5. US 4,704,686 A (Aldinger) and US 6,696,073 B2 (Boyce) — family 4. The foundational "image the patient → model → manufacture the custom implant" references, cited on both the parent and the sibling.
6. US 2018/0250075 A1 (Cho) — family 1 + robotic dependent claims.

Likely § 103 combinations (more realistic than anticipation):

• US 2018/0233222 A1 + Majdouline + US 8,644,568 — planning feedback loop + spine optimization + patient-specific biomechanical scoring.
• US 10,517,681 B2 + US 2017/0035514 A1 — AI surgical guidance + clinical decision support.
• US 8,556,958 / 8,781,557 / 9,198,678 (Conformis line) + US 10,089,413 (Materialise) — patient-adapted implant design + accuracy/stability scoring iteration.
• US 2018/0233222 A1 + US 8,644,568 + additively-manufactured-implant art (US 10,751,188; CN 110575289) — plan → patient-specific implant → print.

On the parent reference: US 10,902,944 B1 appears in the list and is '938's own parent. It is not ordinary § 102 prior art (same inventors, same assignee, and '938 claims its priority); its only role would be as a § 102(a)(2)/§ 102(e)-era or obviousness-type double-patenting reference in limited circumstances. I flag this because an "anticipation by the parent" theory is a common error.

6. Items in the citation list that are NOT prior art to '938 (must be excluded)

7. Near-miss numbers excluded from this analysis

Per the directive to report only the specific number: US 11,389,038 (Omachron; IPR2024‑00938), US 11,487,138 (Luxottica; IPR2025‑00216), US 8,498,388 (Lab Technology reexam 90/019,749), US 8,877,938 / US 9,820,938 (pharma), US 7,817,938 (Canon), and the Patexia "Application No. 11,678,938" record that resolves to a "Deep Frying Apparatus Having a Rotating Basket" — all are different patents/records and none is cited here. Note that US 11,376,076 B2 and US 10,902,944 B1 are used above only as family siblings/parent whose front pages I can read — they are not substitutes for '938, and no prior-art reference is attributed to them as if it were cited in '938 unless it also appears in the '938 citation graph.

8. Confidence and outstanding work

High confidence: the existence and identity of the references listed in § 2, including their numbers and printed dates; the classification of the design patents (§ 2C) and Carlsmed's own later publications (§ 6) as non-anticipatory; the identification of Majdouline et al. (2012), US 2018/0233222 A1, US 10,517,681 B2, and US 8,644,568 B1 as the most § 102-relevant items in the retrievable set.

Cannot confirm / open:

1. The verbatim front-page "References Cited" list for '938 itself. I reconstructed it from the Unified Patents graph plus the sibling and parent front pages. There will be references on '938's face that are not on '076's face (particularly art cited between 2022 and 2023). The definitive source is the front page of the '938 patent PDF (patentimages / Google Patents) or the file wrapper via USPTO Patent Center (https://patentcenter.uspto.gov/), application 17/838,727. That is the first thing to pull.
2. The issued claim text. Without it, the § 102 mapping in § 4 is feature-based, not claim-numbered. No claim-number citation in this document should be treated as verbatim.
3. Whether any given reference anticipates or merely renders obvious. Anticipation requires every element in a single reference; the candidates in § 5 differ on the "reference-patient outcome data" element, which is the apparent point of novelty. I have deliberately not asserted anticipation where the element match is unverified.

Recommended verification links: USPTO Patent Center — https://patentcenter.uspto.gov/ · USPTO Patent Public Search — https://ppubs.uspto.gov/ · Google Patents — https://patents.google.com/patent/US11678938/en · Unified Patents — https://portal.unifiedpatents.com/patents/patent/US-11678938-B2 · WIPO/PCT for PCT/US19/50885 ISR.

Sources: Google Patents record for US 11,678,938 and the sibling/parent documents US 11,376,076 B2 (patentimages PDF), US 11,806,241 B1, US 11,432,943 B2, US 10,902,944 B1, US 2021/0382457 A1, US 2023/0034731 A1; Unified Patents portal page for US‑11678938‑B2 ("Patent Art (286)"); PubChem patent record for US‑11678938‑B2; freepatentsonline record of US 2022/0313362 A1; Carlsmed S-1/registration statement (aprevo technology platform patent table, showing app 17/838,727 → US 11,678,938).

I'll try to close the single biggest gap carried through every prior section — the missing claim text — before running the § 103 analysis.

Let me make two more attempts to obtain the actual issued claim language, since that is the gating fact for any § 103 analysis.

Obviousness Analysis Under 35 U.S.C. § 103 — US 11,678,938 B2

Patent: "Patient-specific medical systems, devices, and methods" · Carlsmed, Inc. · App. 17/838,727 · Continuation of 17/342,439 (now US 11,376,076 B2) → PCT/US21/12065 (filed 2021-01-04) → CIP of 17/124,822 (now US 11,854,683 B2), filed 2020-12-17, and CIP of 16/735,222 (now US 10,902,944 B1), filed 2020-01-06.

0. What I did and did not obtain in this session

I ran four additional targeted retrieval attempts specifically to close the claim-text gap that every prior section flagged. I still did not obtain the issued claim text for '938. What I did obtain is new and worth recording:

• '938's own pre-grant publication, US 2022/0313362 A1 (app 17/838,727), lists 26 claims. The freepatentsonline record renders the claim numbers (1–26) but not the claim bodies. So the family's first-filed claim set was 26 claims; the issued patent's claim count may differ and I cannot confirm it.
• That same record shows the pre-grant publication cited only four US patent references on its face (US 2018/0303552 A1; US 2017/0252107 A1; US 2012/0150243 A1; US 2007/0276501 A1) — i.e., the heavy prior-art citation load on the granted patent was added during prosecution, consistent with the ~286-reference graph discussed earlier.
• The continuation chain above is now confirmed from the applicant's own cross-reference paragraph; this matters for § 103 timing (see § 2).

Consequence for this document: the analysis below is an element-based obviousness assessment, not a claim chart. I do not quote or number claims. Where the outcome of the analysis turns on claim scope I say so rather than assume it.

Date conflict re-flagged (carried forward, unresolved): prior sections record "today" as 2026-09-17 and the Carlsmed v. DeGen Medical suit as filed 2026-09-15; this task states today is 2026-04-26. This does not affect the § 103 analysis (all prior art predates the 2020–2021 filing window), but it does affect any § 315(b) petition calculus, and I will not silently pick one.

1. Legal framework applied

I analyze under Graham v. John Deere as refined by KSR Int'l v. Teleflex, 550 U.S. 398 (2007), using the enumerated rationales: (A) predictable use of prior-art elements according to their established functions; (B) known technique to improve a similar device/method; (C) a finite number of identified, predictable solutions ("obvious to try"); (D) design incentives and market forces; (E) express teaching/suggestion/motivation in the references; (F) design considerations dominant at the time.

The single most important structural point about this patent for § 103 purposes: the independent claims are, on their face, data-processing claims. The abstract's four verbs — receive, compare, select, generate — are the same four verbs that clinical decision-support, case-based-reasoning, and patient-similarity-retrieval systems had been executing for two decades. The asserted novelty resides in which data is compared and why the subset is chosen, not in any new machine, imaging modality, algorithm class, or material. That is the posture in which § 103 is hardest to defend and easiest to prosecute to allowance (which is, in fact, what happened).

The patent's own Background section is an evidentiary gift to an obviousness challenge. It concedes: "although digital data collection and processing power have improved, technologies using collected data to determine optimal treatment protocols have lagged" and that "conventional technologies in the field of orthopedics may lack the capability to draw upon large data sets." That is a judicial admission of the known problem and the known direction of solution — precisely the material that supplies the motivation prong under KSR factor (D)/(E).

2. The critical-date problem — a threshold § 103 weapon

Before reaching combinations, note that the 2020-01-06 priority date is contestable, and the contest is asymmetric.

The chain runs through PCT/US21/12065, filed 2021-01-04, which the applicant describes as a continuation-in-part of both 16/735,222 (2020-01-06) and 17/124,822 (2020-12-17). A CIP by definition adds matter. [CL ※] (a) The pre-grant publication of the '938 family is US 2022/0313362 A1 (this was under review in prior sections). (b) Its chain includes US 10,902,944 B1 ('944) and US 11,376,076 B2 ('076). (c) Its continuation siblings include US 12,137,983 B2 ('983) and US 12,376,907 B2 ('907).

Consequences:

1. If any issued claim of '938 is not supported under § 112 by 16/735,222, its effective filing date shifts to 2020-12-17 (17/124,822) or 2021-01-04 (the PCT). That opens an intervening-art window of roughly 2020-01-07 to 2021-01-04 that the earlier prior-art section correctly closed off on the assumption of a 2020-01-06 date.
2. Art previously excluded as post-dating priority must be re-examined. The prior-art section excluded CN 111281613 A (prio. 2020-02-15, 3D-printed bionic porous artificial vertebral body) as pre-dated by the 2020-01-06 date. On a 2021-01-04 effective filing date it is inside the window as a potential § 102(a)(1) printed publication. Same re-examination applies to any competitor publication from 2020.
3. This is a § 102 and § 103 argument simultaneously: intervening art is prior art under § 102(a)(1)/(a)(2) and is available for combination.
4. The chain also confirms that the family is a large, near-identical set — which raises an obviousness-type double patenting exposure independent of § 103 (see § 8).

I do not assert that priority fails; I assert that the applicant's own CIP-based chain makes the 2020-01-06 date an assumption rather than a settled fact, and that resolving it is a precondition to any serious § 103 opinion.

3. Person of ordinary skill in the art (PHOSITA)

For a 2020–2021 effective filing date, a reasonable formulation is: a computational biomechanics / surgical-planning engineer with (i) a graduate degree (or equivalent experience) in biomedical, mechanical, or software engineering, (ii) 2–4 years building medical image processing and/or surgical planning or medical-device CAD software, and (iii) collaboration with, or literacy in, spine surgery metrics (lordosis, Cobb angle, pelvic incidence, sagittal vertical axis). Alternatively, a spine surgeon with programming/planning-tool experience. Critically, PHOSITA is the same person for all four claim families — the software and the implant design live on one workbench. That identity of the skilled person is what makes the combinations below unremarkable rather than "unpredictable."

4. Element decomposition — the obviousness skeleton

Using the four claim families identified in the earlier summary section (never numbered as claims):

The only element with genuine § 103 risk for a challenger is E3(b) — outcome-based filtering of reference patients. Everything else in this skeleton is squarely in the prior art. I therefore build combinations that attack E3(b) head-on and then explain why the remaining elements fall.

5. § 103 combinations

Combination 1 — The core method (E1–E4): similarity engine + plan generator + spine optimizer

References: US 2018/0233222 A1 (Mako Surgical — planning with multiple feedback loops) + Majdouline et al., Scoliosis 7:21 (2012) + US 11,189,367 B2 (Canon Medical Systems — "Similarity Determining Apparatus and Method," prio. 2018-05-30) + US 8,644,568 B1 (O.N. Diagnostics).

Sourcing flag: US 11,189,367 B2 surfaced in this session's searches, not in the earlier prior-art section. It is not confirmed on the '938 face. It is included because it is the closest art I have found on E2/E3(a) and its absence from the face would itself be notable. Verify before use.

Mapping. Canon '367 discloses storing representations of a plurality of patient data sets generated by applying a model with respect to a plurality of features, obtaining a representation of a further medical data set, and running a similarity-determination process between them — that is E2 plus the E5 similarity score, verbatim in substance, and it is agnostic as to disease, so it reads on spinal pathology data. Mako '222 discloses a surgical planning engine that generates candidate plans and iteratively evaluates them against objectives with multiple feedback loops — that is E4, and its "objectives" lexicon supplies the evaluative/selection concept. Majdouline supplies spine-specificity and optimization: computer simulation and optimization of instrumentation strategies for adolescent idiopathic scoliosis — the same clinical domain, the same physiologic target (post-operative alignment), and an explicit optimization step. O.N. Diagnostics '568 supplies automated patient-specific biomechanical evaluation of a bone-implant construct.

Motivation (KSR factors A, C, E). All four references operate in one field of endeavor — computer-assisted surgical planning/decision support — and address one problem (which implant/plan for this anatomy). Combining a similarity engine with a candidate-generator/evaluator is the predictable use of four known elements, each performing exactly the function it was designed to perform, with the only "new" result being a digital output. Majdouline's own optimization step teaches away from nothing and expressly motivates the extension: once you can score candidate strategies against a simulated outcome, applying that scoring to a cohort rather than a single patient is the routine next step. The finite-solution rationale (KSR factor C) applies with unusual force: a planner has a small, identified menu of options (fusion vs. non-fusion; anterior vs. posterior; decompress vs. not; level count), so "try the ones that worked in similar patients" is the paradigmatic obvious-to-try.

Reasonable expectation of success. High. Nothing in E1–E4 requires the plan to actually succeed, or requires any physiologic unpredictability to be overcome. The operation is a computation over a data structure. A PHOSITA would expect the combination to produce the claimed output essentially every time.

The E3(b) challenge. Neither Canon '367 nor Mako nor Majdouline, standing alone, filters reference patients by their realized outcomes. That gap is closed by (i) Majdouline's optimization objective (a predicted-outcome filter over strategies, from which the cohort-level analogue is a routine extension), and (ii) the ubiquitousness of outcome-linked spine data by 2019 — the specification itself recites HRQL, fusion status, activity level, and complications as ordinary registry fields, and the earlier section's US 2017/0035514 A1 (Abbott — clinical decision support) and US 2011/0196451 A1 (Medtronic — distributed medical data) show the aggregation architecture. For a cleaner single-reference hit on "retrieve similar patient cases and then retrieve the treatment given," US 9,997,260 ("Retrieval of similar patient cases based on disease probability vectors") is striking — its claim 9 expressly recites retrieving and presenting treatment information for retrieved similar cases — but I could not confirm it is on the '938 face, and I flag it as a lead to verify, not as a settled citation.

Combination 2 — The implant-design family (family 4): patient-matched design chain + scoring + additive manufacturing

References: US 8,556,958 B2 and the Conformis chain (US 8,781,557; US 9,198,678) + US 10,089,413 B2 (Materialise — accuracy maps and stability analysis) + US 8,644,568 B1 + US 10,751,188 B2 (additively manufactured patient-specific sacral prosthesis) and/or CN 110575289 A (3D-printed patient-specific bone prosthesis), optionally with US 11,189,367 B2 for the selection step.

Mapping. E4 (device design) is the entire subject matter of the Bojarski/Conformis line ("patient-adapted and improved orthopedic implants, designs and related tools"). Materialise '413 supplies the design-and-iterate-against-criteria step (accuracy/stability scoring). '568 supplies patient-specific biomechanical computation. The AM references supply E7 fabrication. '367 supplies E2/E3(a).

Motivation (factors A, B, D, F). The field's dominant design consideration by 2019 was fit to the individual patient — explicitly recited as the problem in the '938 Background. Factor B (improving a known device by a known technique) is met: additively manufacturing a shape derived from patient imaging was a well-known technique by 2019. Combining a patient-similarity engine with a patient-matched design pipeline yields no more than the predictable sum of the parts — which is the KSR definition of obviousness, not an invention.

Combination 3 — Machine-learning dependent subject matter (E11, and any ML-based claim)

References: US 10,517,681 B2 (Intuitive Surgical — AI guidance for robotic surgery) + US 2018/0233222 A1 (Mako) + US 2017/0035514 A1 (Abbott).

Motivation (A, C). Once the receive/compare/select/generate architecture of Combination 1 is in place, implementing the "compare" step with a trained model is not an inventive step; it is a substitution of one known computational tool for another. KSR is explicit that "a court must ask whether the improvement is more than the predictable use of prior art elements according to their established functions," and where the reference (Intuitive '681) already uses AI to recommend a surgical course, the substitution is predictable. Competitive pressure to improve predictive accuracy (factor D) supplies motive.

Combination 4 — Robotic execution (E8)

References: US 2018/0250075 A1 (Cho — architecture for developing and robotically performing a medical procedure) + US 2018/0116727 A1 (RIT — guided pedicle screw placement) + US 2018/0233222 A1 (Mako, itself a robotic-surgery company).

Motivation (A, D). Generating control instructions to cause a robot to execute an approved surgical plan is the express purpose of each of these references. Where the plan already exists in machine-readable form, automating its execution is a known technique applied to a known structure, with a strong commercial incentive (Mako, Intuitive, Medtronic Mazor all marketed the plan-to-robot pipeline).

Combination 5 — Multi-level / multi-implant (E10)

References: WO 2019/165152 A1 (expandable spinal implant) + US 2016/0354213 A1 (implant/graft for adjacent vertebral bodies) + US 10,631,988 B2 (posterior sacroiliac fusion implant) + the Conformis/Bojarski design chain.

Motivation (A, D). Multilevel lumbar fusion (e.g., L2–L5) is standard surgical practice; constructing a distinct, patient-matched interbody for each treated level is the application of a known design technique to each of a known set of locations. The recited "different geometries" at different levels is a result, not an inventive step — anatomic variation at L2–L3 versus L4–L5 is inherent to the patient, not contributed by the inventor.

Combination 6 — Provider/resource parameters (E9)

References: US 2017/0035514 A1 (Abbott) + US 2014/0100886 A1 (DePuy Synthes — orthopedic implant inventory management) + US 2011/0196451 A1 (Medtronic — distributed medical data network).

Motivation (A, D). Filtering a recommendation engine by institution/surgeon/device-availability variables is the ordinary operation of any medical data-processing system and is taught by the inventory-management and distributed-network references. Weighting and threshold selection (E5) is routine optimization of a known parameter, which KSR and In re Aller treat as obvious absent a showing of criticality.

Combination 7 — GUI/display subject matter

The design patents (D420,995; D436,580; D548,242; D614,191; D653,514; D656,153; D774,076; D797,760; D798,894; D845,973; D845,974; D872,117; D896,825; D916,868; D921,675; D938,461) contribute nothing to § 103 for method/system claims. I state this to correct a common error: a large design-patent citation block is a count, not a validity theory, and a challenger who presents design patents as § 103 art for a functional claim will lose credibility on the rest of the petition.

6. Why the (probably) narrowest claim is still vulnerable — the E3(b) counterargument and its rebuttal

Patent owner's best argument: No single reference and no prima facie combination teaches selecting reference patients by both similarity and their realized treatment outcome, and then using that doubly-filtered cohort to generate a device design. The specificity of the two-axis filter plus the design output is the point of novelty, and the PTO allowed over 286 references — evidence that the examiner could not find the combination.

Why that argument is weak:

1. Citation count ≠ patentability. The prior-art section already established that the 286-item list is a citation graph, not the examiner's considered art, and includes the patent's own later publications and post-priority foreign filings. A large count proves nothing.
2. The two axes are individually known and their combination is a predictable design choice. Similarity retrieval is Canon '367 (and US 9,997,260 as a lead). Outcome-linked data is standard in spine registries and is recited by the patent itself as ordinary. Combining two known filters to improve a recommendation engine is a known technique used to improve a similar device (KSR factor B) — indeed, filtering by outcome is the whole reason one would retrieve similar cases in the first place. Motivation is not merely available; it is inherent in the purpose of a case-based-reasoning system.
3. There is no teaching away, no criticality, and no unexpected result on the record. Absent a showing that the two-axis filter produces a result that a PHOSITA would not have predicted, the combination is obvious as a matter of law.
4. The output limitation does not rescue the claim. Outputting a design instead of a recommendation is a change in output format, not in inventive concept. For § 103 the question is whether the claim as a whole would have been obvious, and "compute a design from the filtered cohort" is the same computation with a different downstream consumer.
5. The strongest real obstacle is evidentiary, not legal: the challenger must produce the actual claim text to know whether E3(b) is even recited in independent form, or only in a dependent claim. If it is dependent-only, the independent claims are on their face anticipated or obvious on Combination 1 alone, and the dependent claim is attacked with the outcome-linked-data art. This is why claim retrieval remains the gating step — a point all prior sections made and which this session could not resolve.

7. Rebuttal evidence the patent owner will (and will not) have

Available pro-patent evidence: commercial success of the aprevo patient-specific implant line; a recently public issuer (Nasdaq: CARL per prior sections); the September 2026 assertion against DeGen Medical; and the narrative that patient-specific spine implants were an unmet clinical need.

Why it likely fails:

• No nexus. The asserted claims, on their face, are directed to generating a plan/design, not to the implant that is sold. Commercial success in implants does not presumptively flow from a data-processing claim. Any nexus argument must be built on the actual claim text — again unavailable.
• The "unmet need" cuts the other way. The '938 Background admits the need was known, and the earlier prior-art section shows the field was crowded with patient-specific design solutions from 1987 onward. Long-felt need without failure-of-others is weak.
• No adjudicated validity. The prior PTAB section is correct that there are zero AIA proceedings. There is therefore no PTAB or court record of the patent owner having overcome any § 103 combination, and no § 315(e)(2) estoppel limiting anyone.

8. Adjacent risks the § 103 analysis exposes (flagged, not developed)

• Obviousness-type double patenting. The family now includes US 11,376,076; US 11,854,683; US 12,137,983; US 12,376,907; and US 2026/0000460 A1, all sharing one abstract and specification. If any issued claim is not patentably distinct from another, OTDP is an independent invalidity/unenforceability exposure requiring a terminal disclaimer — and a terminal disclaimer would cap enforceable term.
• § 112. The claims appear to be functionally recited ("generate… based on the selected subset") while the specification discloses the algorithm only at the level of "assign values and aggregate." That is a written-description/enablement vulnerability independent of § 103.
• § 101. Receive/compare/select/generate over a patient database is close to the Mayo/Alice core. The saving grace would be the manufactured-implant output — but that is exactly the distinction that does not defeat § 103.
• Timing gates. As prior sections established, no IPR/PGR exists; the § 315(b) one-year clock runs only if '938 is among the eight patents asserted in the (reported) 2026-09-15 D.S.C. complaint, which is still unconfirmed from public sources.

9. Bottom line

On the claim families as I can reconstruct them, US 11,678,938 B2 is obvious under § 103. The core independent subject matter (E1–E4) is met by a three-reference combination drawn almost entirely from the patent's own cited art — US 2018/0233222 A1 (plan generation/evaluation) + Majdouline et al. 2012 (spine-specific optimization) + US 8,644,568 B1 (patient-specific biomechanical computation) — with the similarity-selection step supplied by US 11,189,367 B2 (verify) and/or US 9,997,260 (verify). Every dependent family (ML, robotic control, multi-level implants, provider/resource filtering, scoring/thresholding) is met by a routine, motivation-supported combination of art in the same field. The 286-reference count is not a validity shield.

The three things that would change this conclusion — all unresolved:

1. The issued claim text. Not obtained in this session despite four targeted attempts; the pre-grant publication lists 26 claims. Without it, this is an element-level assessment, not a chart. The prior sections' standing caution applies unchanged and I do not quote claim numbers.
2. The effective filing date. The applicant's own CIP-based chain (through PCT/US21/12065, filed 2021-01-04) makes the 2020-01-06 date contestable and, if lost, opens an intervening-art window that revives references previously excluded (e.g., CN 111281613 A).
3. Whether US 11,189,367 B2 and US 9,997,260 are on the '938 face. These are my strongest E2/E3(a) hits and both are newly surfaced, unfiled leads.

Confidence: High that the independent-claim architecture is vulnerable to a well-constructed § 103 combination in the general-purpose clinical-decision-support and patient-matched-implant design art. Low-to-moderate that any specific combination anticipates (as opposed to renders obvious) — the E3(b) two-axis filter is the likely point of novelty and is best attacked under KSR factor B plus routine-optimization, not anticipation. Explicit uncertainty: I have not verified the presence of US 11,189,367 B2 or US 9,997,260 on the '938 face; I have not read the issued claims; and I have not resolved the priority-date question.

Verification links: https://patents.google.com/patent/US11678938/en · https://patentcenter.uspto.gov/ (app 17/838,727) · https://www.freepatentsonline.com/y2022/0313362.html (pre-grant pub., 26 claims listed) · https://portal.unifiedpatents.com/patents/patent/US-[11678938](/patent/11678938)-B2

Sources for this section: US 2022/0313362 A1 (freepatentsonline) — cross-reference paragraph establishing the CIP chain and the 26-claim count; US 2018/0233222 A1; Majdouline et al., Scoliosis 7:21 (2012); US 10,517,681 B2; US 8,644,568 B1; US 10,089,413 B2; US 4,704,686 A; US 6,696,073 B2; US 8,556,958 B2; US 2017/0035514 A1; US 2018/0250075 A1; US 2018/0116727 A1; US 2014/0100886 A1; US 2011/0196451 A1; WO 2019/165152 A1; US 10,751,188 B2; CN 110575289 A; CN 111281613 A; US 11,189,367 B2 (Canon Medical Systems); US 9,997,260 (retrieval of similar patient cases) — the last two newly surfaced this session and not confirmed as cited on the '938 face.