US 5287272

I'll search for authoritative records on this patent number before summarizing.

Let me pull the claim text and check litigation/docket records.

US 5,287,272 — Analyst Summary

Sources consulted (fetched 2026‑04‑26): the patent full text you supplied (authoritative, from https://patents.google.com/patent/US5287272/en), plus live searches returning https://FreePatentsOnline.com/[5287272](/patent/5287272).html, https://uspto.report/patent/grant/5287272, and the CIP/priority record at https://patentimages.storage.googleapis.com/.../WO1997041416A1.pdf.

1. Bibliographic data

⚠️ Title caveat: the parent application's European counterpart (EP 0 336 608, filed 1989‑03‑22) is titled "Neural network based automated cytological specimen classification system and method." Both titles appear in circulation for this family; for US 5,287,272 specifically, both the Google Patents header and the FreePatentsOnline record give the title without the "neural network based" prefix.

2. Abstract (verbatim)

"An automated screening system and method for cytological specimen classification in which a neural network is utilized in performance of the classification function. Also included is an automated microscope and associated image processing circuitry."

3. Plain-language overview

The patent covers an automated Pap‑smear screening pipeline built around two chained classifiers plus a human backstop:

1. Low‑resolution pass — the automated microscope (5X objective) scans 2 mm × 2 mm "macro‑tiles" along a serpentine path, subdivides each into sixteen "tiles," and builds (a) a sharpness image for a focus map, and (b) a gray‑scale/histogram measure to decide which tiles actually contain cellular material. Illumination, autoiris, and objective carriage are motor‑controlled.
2. High‑resolution pass — a 20X objective re‑images only the tiles of interest, using the stored focus map for fast refocusing.
3. Primary classification (low‑level feature extraction) — adaptive thresholding, down‑sampling, then erosion and non‑connecting dilation to break up overlapping cells, followed by integrated optical density (IOD) screening. This is claimed to reduce ~1,000 objects in a tile image to roughly ten candidates, and it inherently centers each surviving object on its dark nuclear centroid.
4. Secondary classification (adaptive / neural net) — a three‑layer back‑propagation network (implemented on an HNC Anza Plus board, with a PIPE image processor from ASPEX) is fed 24×24‑pixel images pre‑centered by the primary classifier. Trained to output 0.1 for benign and 0.9 for non‑benign; a 0.65 cut‑off is used in the prototype. The spec reports >80% of benign images filtered out.
5. Tertiary classification (human) — up to ~64 suspect cells at a time are displayed as 128×128‑pixel color tiles on a high‑resolution monitor for a cytotechnologist/cytologist, who can tag cells, recall them, and physically mark the slide (a ~0.25 mm ink dot via the slide marker).
6. Workflow hardware — robotic slide handler (vacuum‑held "foot," cassette elevator), barcode reader for specimen authorization and report correlation, motorized X‑Y stage with position encoder, and slide marker.

The stated central technical insight is that combining feature extraction with template matching removes the need for image segmentation and removes the classic neural‑network "where is the digit on the envelope" problem: the primary classifier's IOD/erosion‑dilation filter automatically centers the suspect nucleus in the network's input array.

4. Independent claims (plain language)

I retrieved the claim set via FreePatentsOnline (https://FreePatentsOnline.com/5287272.html). The record appears to contain at least 26 claims; my retrieved text is truncated after claim 25, and claim 25 is dependent. So the independent claims I can verify are 1, 23, and 24. Note all apparatus claims use "means for" language, implicating 35 U.S.C. § 112(f).

Claim 1 — the core apparatus claim. A cytological specimen classifier comprising: (a) microscope means for viewing at least part of a specimen; (b) camera means for imaging that view; (c) image digitizing means for producing a digital representation; (d) a primary classifier identifying objects in the digital representation based on a detectable feature; and (e) secondary adaptive classifier means for recognizing cells whose patterns are atypical of patterns expected in the specimen, among the objects the primary classifier identified. In short: microscope + camera + digitizer + rule‑based pre‑filter + adaptive (learned) second‑stage classifier.

Claim 23 — the "quality control" apparatus claim. A cytological specimen classifier specifically for classifying specimens previously screened as not containing pre‑malignant or malignant cells, comprising the same microscope/camera/digitizer/primary‑classifier chain, then (e) secondary adaptive classifier means for recognizing cells with patterns typical of pre‑malignant or malignant cells among the prescreened objects, and (f) a tertiary classifier for detecting malignant and pre‑malignant cells among what the secondary classifier flags. In short: a re‑screen/rescreen machine for already‑read negative slides, with a mandatory human (or equivalent) third stage.

Claim 24 — the core method claim. A method of classifying objects in a cytological specimen: (a) obtain a view; (b) create an image of it; (c) produce a digital representation; (d) primarily classify objects in the digital representation based on a detectable feature; and (e) secondarily classify, using adaptive processing, cells having features atypical of cells expected in the specimen, among the objects identified in the primary step. In short: the method counterpart of claim 1 — digitize, pre‑filter morphologically, then apply an adaptive/learned classifier.

Selected dependent claims that sharpen scope: cl. 2 (tertiary classifier), cl. 3 (high‑resolution color monitor), cl. 4 (secondary classifier = neural network), cl. 5–7 (slide information reader; barcode reader; OCR), cl. 8 (slide marking of recognized cell locations), cl. 11 (patterns typical of malignant/pre‑malignant cells), cl. 12 (patterns typical of endometrial tissue), cl. 13 (primary classifier erodes and dilates), cl. 14 (color‑based screening), cl. 15 (parallel processor), cl. 19 (primary statistical classifier), cl. 21 (threshold integrated optical density), cl. 25 (tertiary classification step).

5. Prosecution / post‑grant events

• A publication US5287272B1 dated 1996‑08‑27 appears in the Patent Center–derived record, consistent with a post‑grant (ex parte reexamination) certificate. ⚠️ I could not verify the content of that certificate (which claims were confirmed, amended, or added) from the sources retrieved; treat any claim‑scope statement as subject to that certificate.
• Statutory term: 17 years from the 1994‑02‑15 issue date would have run to 2011‑02‑15, but the record shows anticipated expiration 2013‑08‑27, consistent with additional term (e.g., § 154(b) adjustment). The patent is expired, so it cannot be asserted prospectively.

6. Litigation / CAFC 2026 docket check — findings and uncertainty

• I found no CAFC 2026 docket entry, opinion, or appeal naming U.S. 5,287,272. My searches for the number in Federal Circuit and docket contexts returned unrelated 2026 decisions (Global Tubing v. Tenaris, Magnolia Medical v. Kurin, AGI v. Farmers Edge, Ironburg v. Valve, Gamevice/Nintendo) and general appellate‑practice material — none cite this patent. I also did not find any docket record naming it as an asserted patent in any year.
• Limitation to flag: my searches hit public web/patent aggregators, not PACER, the CAFC's own docket system, or the USPTO Patent Center/PTAB APIs directly. Absence of a hit in my results is not proof that no such docket exists. Given the patent expired in 2013, a 2026 appeal on it would be unusual.
• One district‑court decision I did surface is a 2005 Markman memorandum in Cytyc Corporation v. TriPath Imaging, Inc. (https://ipmall.info/sites/default/files/hosted_resources/Markman/pdfFiles/2005.11.28_CYTYC_CORPORATION_v._TRIPATH_IMAGING_INC.pdf), which construes U.S. 6,327,377 and U.S. 5,257,182 — patents in/adjacent to this family (AutoCyte/TriPath–Cytyc dispute over the ThinPrep vs. AutoCyte/Papnet technology). ⚠️ I could not confirm from the retrieved text that the '272 patent itself was asserted in that case; the snippet names only the '377 and '182 patents. Do not treat the '272 as a litigated patent without verifying that docket directly.

7. Family context (useful for validity/priority work)

Parent US 4,965,725 (Ser. No. 179,060, filed 1988‑04‑08; the "Neural network based automated cytological specimen classification system and method" disclosure) → this CIP, US 5,287,272 → continuations US 5,544,650, US 5,740,270, US 5,939,278, US 6,327,377. Sibling US 5,257,182 (Luck, "Morphological classification system and method") pursues the low‑resolution/high‑resolution rescan and gallery approach. Foreign counterpart EP 0 336 608 A2/A3 (priority 1988‑04‑08), examination citing Tien et al. 1987 and Greenberg US 3,333,248 as "Y" references against claims 1–3, 8–11.

Bottom line

US 5,287,272 (Neuromedical Systems; now TriPath Imaging; expired 2013‑08‑27) claims a staged cytological classifier — digitizing microscope/camera front end, a morphological primary filter (erosion/dilation + integrated optical density), and an adaptive (neural‑network) secondary classifier, optionally with a human/tertiary stage, plus a dedicated quality‑control rescreen variant (claim 23) for re‑reading already‑negative slides. Three independent claims are verified (1, 23, 24); the claim set appears to extend past claim 25, and a 1996 reexamination certificate (US5287272B1) may affect the operative claim text — I could not verify that certificate's contents. No CAFC 2026 docket activity involving this patent was found in my searches.

I'll research litigation involving US 5,287,272. Let me search several sources.

Let me search more specifically for litigation tied to this patent's owners and technology.

Let me dig into the specific cases and verify whether '272 was asserted.

I'll start by checking the canonical structured source and then probing the web for any AIA trial activity on this patent.

The first pass surfaced only unrelated "′528" patents (ParkerVision's 9,118,528) and foreign/irrelevant hits. Let me run targeted searches on the patent owner and litigation lineage.

Let me run final targeted checks on PTAB activity and the patent's term/assertion history.

Proceedings overview

Total AIA trial proceedings on US 5,287,272: 0 — zero IPRs, zero PGRs, zero CBMs (0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution denials), which means no claim of the '272 has ever been canceled or even tested at the PTAB, and there is no § 315(e)(2) estoppel to work with; the defensive posture for a defendant therefore does not rest on IPR outcomes at all — it rests on the fact that the patent is Expired – Lifetime (anticipated expiration 2013-08-27), so any demand letter citing it is litigating past damages under 35 U.S.C. § 286, not infringement of an enforceable right.

Source of the count. The structured "PTAB proceedings on file" block supplied in this prompt is the canonical list from the USPTO Open Data Portal and returns no AIA trial proceedings. I independently searched the web (Google Patents, PTAB/PTACTS document hits, CourtListener, litigation news) on 2026-09-27 and found no IPR, PGR, or CBM naming U.S. Patent 5,287,272. Caution on search noise: heavy search traffic exists for a different patent — U.S. 9,118,528 ("the '528 patent," ParkerVision v. NXP/TI/MediaTek, several IPRs) — and for U.S. 5,658,727 (Medtronic v. Boston Scientific). Neither is this patent. I am not auto-correcting or conflating those.

No AIA trial proceedings to detail

Because the canonical list is empty, there is no proceeding number, panel, institution decision, FWD, settlement, or CAFC appeal to report for the '272. I will not manufacture one. Two structural reasons the absence is credible rather than a data gap:

• Timing. The first AIA petition was filed 2013-09-16; the '272's term ended 2013-08-27 — i.e., the patent was already expired before the first AIA trial was ever instituted. There was no commercial window in which a competitor needed an IPR to clear the patent.
• Procedural fit. PGR was unavailable (effective filing date long predates 2013-03-16); CBM would not have covered this subject matter (image-analysis/cytology apparatus, not a "financial product or service," and the CBM program sunset on 2020-09-16).

Closest analogues — pre-AIA district court litigation (NOT AIA trials)

These are the only adversarial proceedings I can substantiate, and they are civil actions, not PTAB trials. Include them for context only; they provide no estoppel and no claim cancellations.

• Neuromedical Systems, Inc. v. NeoPath, Inc. — S.D.N.Y., filed 1996-07-15 (patent infringement, unfair competition, false advertising); amended complaint 1998-07-27 narrowed to patent infringement only. In the companion preliminary-injunction motion, NSI relied on claim 19 of U.S. 4,965,725 (the '725 patent) and claim 24 of the '272 patent, arguing NeoPath's AutoPap 300 QC system directly infringed those claims. The S.D.N.Y. denied the preliminary injunction in May 1998, finding NSI had not shown a likelihood of success on the merits. The action was removed from the active docket on 1999-04-06 after NSI filed Chapter 11 (1999-03-26).
• NeoPath, Inc. v. Neuromedical Systems, Inc. — W.D. Wash., filed 1997-03-31 (NeoPath asserting three of its own patents against NSI's PAPNET); stayed 1999-03-30 pending NSI's bankruptcy.
• Outcome of the patent estate: NSI's patent estate (including the '272) was sold to AutoCyte in 1999 (with NeoPath acquiring an undivided interest), the litigation was terminated by agreement, and the estate passed to TriPath Imaging, then to Becton Dickinson (2006-12-20). No validity judgment on the '272 ever issued — the cases died with the bankruptcy and the acquisition, not on the merits.
• Later family litigation: Cytyc Corp. v. TriPath Imaging (D. Mass. No. 1:03-cv-11142) litigated the '377 (Luck) and '182 (Luck) patents — not the '272. See the 2007 summary-judgment opinion at govinfo.gov.

Links: US 5,287,272 on Google Patents · full claim set, FreePatentsOnline · 1998 S.D.N.Y. opinion excerpt · D. Mass. Cytyc/TriPath opinion · PTAB E2E portal (definitive negative check): https://ptacts.uspto.gov/ptabweb · CourtListener

Strategic summary

Canceled vs. sustained vs. untested. Nothing is canceled. Because no AIA trial was ever filed, every claim of the '272 stands exactly as issued — including independent claim 1 (cytological specimen classifier: microscope + camera + image digitizing + primary classifier + secondary adaptive classifier means) and independent claims 23 (quality-control classifier) and 24 (method of classifying objects). They are "sustained" only in the trivial sense that no one attacked them at the PTAB; they are best understood as UNTESTED at the Board and EXPIRED at the Office (Google Patents legal status: Expired – Lifetime; anticipated expiration 2013-08-27). Note also the family's terminal-disclaimer structure: related continuations (U.S. 5,544,650; 5,740,270; 5,939,278; 6,327,377) carry terms capped by the family. If a demand letter in 2026 names the '272, the operative defense is the calendar, not prior art.

Estoppel landscape. There is no § 315(e)(2) estoppel — estoppel attaches only to an instituted IPR/PGR, and none exists. Consequently, for a defendant presently facing assertion, every prior-art ground that was ever available remains available (nothing was raised or reasonably could have been raised, because nothing was raised). Practically, however, the more useful bars are statutory rather than estoppel-based:

• § 286 damages lookback (6 years). With the patent expired 2013-08-27, recoverable damages would require suit filed by roughly 2019 against pre-expiration conduct; on a 2026 assertion, the § 286 window has long closed. Past-damages exposure should be the first analytic move.
• No injunctive leverage. An expired patent cannot support prospective injunctive relief.
• Validity is still fair game — with no estoppel and no PTAB record, an invalidity case would be built from scratch, but the classic motivation (clearing an operating product) is absent.

Pattern signals. None of the red flags apply: no serial petitioner, no defensive aggregator (no Unified Patents or similar) involvement found, no PTAB appeal history by the patent owner, and no reexamination/reexam-certificate challenge surfaced. The patent's enforcement history is a 1990s two-front district court war (S.D.N.Y. and W.D. Wash.) against NeoPath that ended in NSI's Chapter 11 and the sale of the estate — not a merits win for the patent owner. Since 1999 the patent has sat in the TriPath/Becton Dickinson portfolio without generating AIA trial activity, and it expired in 2013.

One open item to verify before relying on this: the Google Patents record shows a 1996-08-27 event logged as "Publication of US5287272B1." I cannot confirm from the available materials whether that reflects an ex parte reexamination certificate (which could mean the claims were amended in 1996, in which case the 1997–98 assertion of "claim 24" would reflect post-reexam claim numbering) or merely a bibliographic/publication variant. The structured ODP block and my searches show no reexamination proceeding; treat the B1 designation as an unresolved bibliographic question and pull the certified copy from PatentCenter if claim numbering matters to your case.

Recommended next steps

1. Run the definitive negative on PTAB E2E, not just ODP: https://ptacts.uspto.gov/ptabweb (and the PTAB "Patent Number search" at https://developer.uspto.gov/ptab-api). Record the search date (2026-09-27) and the null result for U.S. 5,287,272 — that null result is your evidence of no PTAB activity, and it's the kind of check worth documenting in a litigation hold memo.
2. If you hold a demand letter citing the '272: lead with expiration. Pull the certified bibliographic data and the expiration calculation from PatentCenter/EPAS; note the Expired – Lifetime status and the 2013-08-27 anticipated expiration shown on Google Patents, then apply the § 286 six-year bar. There is no FWD to cite because no FWD exists — do not let opposing counsel imply otherwise.
3. Check what is actually being asserted. Because there is no enforceable '272, confirm whether the real target is a family member with a later term — U.S. 6,327,377 (filed 1998-04-13), U.S. 5,939,278, U.S. 5,740,270, or U.S. 5,544,650 — or the Luck '182 patent litigated in Cytyc v. TriPath. Those are different patents with different (and possibly longer) terms and different PTAB histories, and Cytyc produced real claim-construction and validity rulings you would want to read.
4. If you nonetheless pursue invalidity (e.g., against a family member), note there is no estoppel constraint and no instituted-IPR prior-art record to build on or be bound by; the 1988–1990 priority date and the Tien/Hecht-Nielsen/Lippmann prior art cited on the face of the '272 define the field.
5. Term-stage milestones: inapplicable — there is no active proceeding, so there is no institution deadline, no oral hearing, and no statutory one-year FWD due date running on this patent.

Caveats: this report reflects the ODP structured record plus web searches conducted 2026-09-27; I found no AIA trial proceeding on U.S. 5,287,272 and have not fabricated a proceeding number, panel, or FWD. Claim-level litigation facts come from the 1998 S.D.N.Y. opinion excerpt (claim 19 of the '725 and claim 24 of the '272) and contemporaneous reporting; the precise docket numbers for the pre-AIA suits and the nature of the 1996-08-27 "B1" event should be confirmed on PatentCenter and PACER/CourtListener before use in a filing.

I'll research the assignment chain before writing anything. Let me pull the available records.

Let me dig for the specific reel/frame records and correspondents.

US 5287272 — Ownership Chain Reconstruction

"Automated cytological specimen classification system and method" · App. 07/420,105 · Filed 1989-10-11 · Issued 1994-02-15 · Anticipated expiration 2013-08-27

Data-gap disclosure up front. I was unable to retrieve reel/frame numbers or correspondents of record for this patent. The USPTO Assignment Center interface was not reachable through the search tooling available to me, and the authoritative file supplied for this patent (Google Patents) reproduces the assignment events from the USPTO reassignment database but does not publish reel/frame or correspondent fields. Everything below is drawn from that authoritative legal-events record plus SEC filings and trade press. No reel/frame or correspondent has been inferred, and no assignment has been invented. Where a field is unavailable I say so rather than filling it. This materially weakens signals 3 and 4 below, which are the fields where the missing data would have been decisive.

Inventors

Nine named inventors on the face of the patent:

Basis for employer: all nine appear as assignors on the three inventor-to-assignee recordations dated 1990-01-05 naming NEUROMEDICAL SYSTEMS, INC. as assignee, with conveyance type "ASSIGNMENT OF ASSIGNORS INTEREST." That is the assignee of record on the issued patent, so it is the employer of record for all nine.

Unusual patterns — none found.

• No inventor retained rights; all nine assigned to the same employer. This is a textbook single-employer corporate assignment, not a fragmented or partial chain.
• No inventor appears to have departed before the chain moved. The departure/fire-sale story here is not an inventor-side event — it is the original assignee's Chapter 11 ten years later (see timeline). The relevant "distress" marker is entity-level, not inventor-level.
• The inventors signed in three grouped recordations on the same date (Hall+Rutenberg / Chaban+Domes+Dulak / Herriman+Knapp+Luck+Porzio). Grouped filings of this kind usually reflect signature logistics, not differing consideration — but I could not verify that, because the underlying assignment documents were not retrievable.

Original assignee

Neuromedical Systems, Inc. ("NSI") — original assignee on the issued patent and named as assignee in the 1990-01-05 inventor assignments.

• Primary line of business: developer and manufacturer of automated cytology screening equipment. Its flagship product was the PAPNET neural-network-based Pap smear screening system — which is the commercial embodiment of this patent family (primary classifier → secondary neural-net classifier → tertiary human classifier, exactly as claimed and described).
• Did they ship a product embodying the claims? Yes. PAPNET was built out and marketed through the early/mid-1990s; the specification's engineering-prototype results (over 80% benign filtration by the secondary classifier; 100,000 cells reduced to 200–250 for human review) are the product's operating parameters. PAPNET never achieved the market penetration NSI needed, and it was NSI's commercial failure that drove the bankruptcy.
• Current status: dissolved. NSI entered Chapter 11 bankruptcy in 1998/1999. Trade reporting (The Dark Report, Volume VI No. 5, April 5, 1999) states: "When Neuromedical Systems entered Chapter 11 bankruptcy, AutoCyte moved rapidly to purchase Neuromedical's patents and other intellectual property." TriPath Imaging's own 10-K confirms the mechanics: "On March 25, 1999, we entered into a purchase and sale agreement to acquire the intellectual property estate of Neuromedical Systems, Inc. … The purchase was completed in May, 1999. This intellectual property estate is held by AutoCyte North Carolina, LLC, our wholly-owned subsidiary."
• Successor chain: NSI → AutoCyte, Inc. (IP estate, 1999) → TriPath Imaging, Inc. (formed September 1999 via merger of AutoCyte + NeoPath) → Becton, Dickinson and Company (acquired TriPath, announced 2006-09-08, closed 2006-12-20, ~$350M).

Assignment timeline

Recorded events as they appear in the authoritative file. Dates are USPTO recording/event dates; executed dates are stated only where independently confirmed. Reel/frame and correspondent are flagged as unavailable on every entry rather than left blank, so the gap is auditable.

1990-01-05 (recorded) — Reel/frame: not retrievable

• Conveyance: Assignment of assignors' interest
• Assignor: Herriman, James M.; Knapp, James; Luck, Randall L.; Porzio, John
• Assignee: Neuromedical Systems, Inc.
• Correspondent: Not retrievable (Assignment Center not reachable; Google Patents legal events omit this field). No recurrence determination possible.
• Context: Formation-stage in-house assignment — founders/employees conveying rights to the operating startup that filed the application.

1990-01-05 (recorded) — Reel/frame: not retrievable

• Conveyance: Assignment of assignors' interest
• Assignor: Hall, Thomas L.; Rutenberg, Mark R.
• Assignee: Neuromedical Systems, Inc.
• Correspondent: Not retrievable.
• Context: Formation-stage in-house assignment (same day, second grouped recordation).

1990-01-05 (recorded) — Reel/frame: not retrievable

• Conveyance: Assignment of assignors' interest
• Assignor: Chaban, Richard; Domes, Robert; Dulak, Thomas
• Assignee: Neuromedical Systems, Inc.
• Correspondent: Not retrievable.
• Context: Formation-stage in-house assignment (same day, third grouped recordation).

1990-12-14 (recorded) — Reel/frame: not retrievable

• Conveyance: Security Interest (record text: "SECURITY INTEREST (SEE DOCUMENT FOR DETAILS)")
• Assignor: Neuromedical Systems, Inc.
• Assignee: Rothschild, Steven B. (natural person, not an entity)
• Correspondent: Not retrievable.
• Context: Securitization / venture-debt collateral — a pledge of NSI's IP as collateral to an individual lender/investor, not a transfer of ownership. This does not move title.

1999-07-26 (recorded) — Reel/frame: not retrievable

• Conveyance: Assignment of assignors' interest ("ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS)")
• Assignor: Neuromedical Systems, Inc.
• Assignee: AutoCyte North Carolina, L.L.C.
• Correspondent: Not retrievable.
• Context: Bankruptcy fire-sale of the IP estate, executed via a purchase and sale agreement dated 1999-03-25 and completed May 1999, recorded 1999-07-26. Buyer was not an NPE — AutoCyte, Inc. was an operating cytology competitor (and later half of TriPath Imaging). AutoCyte NC LLC was, per TriPath's 10-K, a wholly-owned subsidiary holding the estate.

2000-02-28 (recorded) — Reel/frame: not retrievable

• Conveyance: Security Interest
• Assignor: AutoCyte North Carolina, LLC
• Assignee: MMC/GATX Partnership No. 1 and Transamerica Business Credit Corporation
• Correspondent: Not retrievable.
• Context: Securitization / equipment-and-IP financing — third-party lenders taking collateral in the acquired IP estate. Again, a pledge, not a title transfer.

2003-05-09 (recorded) — Reel/frame: not retrievable

• Conveyance: Termination of Security Interest
• Assignor: MMC/GATX Partnership No. 1; Transamerica Business Credit Corporation
• Assignee: AutoCyte North Carolina LLC
• Correspondent: Not retrievable.
• Context: Securitization release — lender lien discharged, confirming the debt was repaid or the facility terminated.

2003-05-12 (recorded) — Reel/frame: not retrievable

• Conveyance: Assignment of assignors' interest
• Assignor: AutoCyte North Carolina, L.L.C.
• Assignee: TriPath Imaging, Inc.
• Correspondent: Not retrievable.
• Context: Internal reorganization / title consolidation — moving the estate from the wholly-owned IP-holding subsidiary up to the listed operating parent. Note the timing: this is the only title transfer in the record with any temporal proximity to litigation (Cytyc's 2003 declaratory-judgment action against TriPath), which is why signal 5 is scored "unclear" rather than "not present."

Date not in record — post-issuance chain (Becton Dickinson)

• No assignment from TriPath Imaging, Inc. to Becton, Dickinson and Company appears in the legal-events record. The 2006 transaction was an all-cash share merger ($9.25/share, ~$350M, announced 2006-09-08, completed 2006-12-20), in which title to subsidiary assets typically passes by operation of law rather than by a separate recorded assignment. Google Patents still lists the current assignee as "TriPath Imaging Inc." This is a real record-keeping artifact worth noting: the recorded chain and the beneficial chain diverge here.

2013-08-27 — Anticipated expiration (legal status: Expired – Lifetime). The patent is expired; it cannot be asserted today by anyone.

Timeline diagram

timeline
    title Ownership of US 5287272
    1988 : Parent application filed
    1989 : CIP application filed
         : Neuromedical Systems owns the rights
    1990 : Inventors assign to Neuromedical
         : Security interest to Rothschild
    1994 : Patent issued
    1998 : Neuromedical files Chapter 11
    1999 : IP estate sold to AutoCyte North Carolina LLC
         : AutoCyte merges with NeoPath into TriPath
    2000 : Security interest to MMC GATX and Transamerica
    2003 : Security interest terminated
         : Estate assigned up to TriPath Imaging
         : Cytyc files suit against TriPath
    2006 : BD acquires TriPath Imaging
    2013 : Patent expires

NPE / troll-pattern signals

1. Shell-entity transfer — NOT PRESENT.
The only LLC in the chain is AutoCyte North Carolina, L.L.C. (recorded 1999-07-26). It has the right shape for the tell — single-purpose IP-holding LLC, "North Carolina" placeholder naming. But the concrete evidence defeats the call: TriPath's SEC filings describe it as "our wholly-owned subsidiary" whose sole function is to hold the IP estate of an acquired operating business. It was never a licensing-only vehicle; the parent shipped PAPNET-lineage and AutoPap/SurePath/FocalPoint product. Per the operating rule, shape alone is not a finding — and here the substance affirmatively contradicts it.

2. Known asserter in the chain — NOT PRESENT.
None of the entities in this chain matches any listed NPE: Neuromedical Systems, Inc.; AutoCyte North Carolina, L.L.C.; TriPath Imaging, Inc.; Becton, Dickinson and Company. Nor do the secured parties (Steven B. Rothschild; MMC/GATX Partnership No. 1; Transamerica Business Credit Corporation) — these are lenders, and a security interest is not ownership. No Acacia, Marathon, IV, IPNav, Wi-LAN, Conversant, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp or Spangenberg entity appears anywhere in the title chain.

3. Repeat correspondent across the chain — UNCLEAR (DATA UNAVAILABLE).
This is the one signal I cannot score, and it is the signal most likely to change the analysis. The correspondent of record is exposed by USPTO Assignment Center per recording, and none of the recordations in this chain surfaced a correspondent through the sources available to me. I therefore cannot test for a recurring attorney/firm across the 1990, 1999, 2000, 2003 entries, nor check any name against Unified Patents / RPX / Patent Progress NPE assertion lists. Flagging as an open item for manual pull at https://assignmentcenter.uspto.gov/ — searchable by patent number, eight records expected. I will not guess.

4. Cascading transfers — NOT PRESENT.
The chain has exactly two substantive title transfers across 14 years: NSI → AutoCyte NC LLC (1999-07-26) and AutoCyte NC LLC → TriPath Imaging (2003-05-12). The intervening 2000-02-28 and 2003-05-09 filings are a security interest and its termination — a matched pair that leaves title untouched. Four years between the transfers, against a "<24 months through chained LLCs" test. Even if the two secured-party entries were miscounted as transfers, the parties share no correspondent address or principals that I could establish.

5. Pre-litigation transfer — UNCLEAR.
The 2003-05-12 assignment up to TriPath Imaging, Inc. sits in the same window as Cytyc's 2003 declaratory-judgment action against TriPath ("claiming that certain of TriPath's patents were invalid and not infringed upon by Cytyc's imaging system," per TriPath's 10-K). A subsidiary-to-parent consolidation on the eve of a patent case is a classic clean-standing move. But (a) the 2003 suit was Cytyc suing TriPath — TriPath was the defendant, so the consolidation was defensive standing hygiene, not an offensive pre-assertion flip; (b) I could not confirm the suit filing date to test the six-month window; and (c) I could not confirm that US 5287272 specifically was a patent-in-suit. Marked unclear on the evidence, not on the pattern.

6. Bankruptcy fire-sale — PRESENT.
Confirmed on the record: Neuromedical Systems, Inc. entered Chapter 11 (reported by The Dark Report, April 1999); the purchase and sale agreement for its IP estate was signed 1999-03-25, the purchase completed May 1999, and the transfer recorded 1999-07-26 as an assignment from NSI to AutoCyte North Carolina, L.L.C. However — and this is the analytically important half — the buyer was an operating competitor, not an NPE. A distressed sale into an operating company is the inverse of the Kodak/Nortel/Polaroid-to-troll pattern this signal is designed to catch.

7. Privateering — NOT PRESENT.
No evidence that NSI transferred the estate to an NPE to assert on its behalf. NSI was the insolvent seller; the buyer (AutoCyte) had its own competing product and its own R&D program. There is no SEC or Patent Progress / EFF reporting of a privateering arrangement in this chain. Direct competitor-versus-competitor litigation (Cytyc v. TriPath, 1999–2001 settlement and 2003 re-filing) is ordinary commercial patent warfare, not privateering.

8. Defensive aggregator (anti-NPE) — NOT PRESENT.
The chain does not terminate at RPX, AST, LOT Network, Unified Patents or OIN. It terminates at Becton, Dickinson and Company, a large operating medical-technology manufacturer that sells BD SurePath and BD FocalPoint products practicing this lineage. The patent was never neutralized by a defensive aggregator; it simply expired on 2013-08-27 while held inside an operating company.

Verdict

Operating-company assertion.

Justification: the title chain runs Neuromedical Systems, Inc. → AutoCyte North Carolina, L.L.C. (1999-07-26, a wholly-owned operating subsidiary, not a licensing vehicle) → TriPath Imaging, Inc. (2003-05-12) → Becton, Dickinson and Company, with the only interruptions being two matched security-interest filings (1990-12-14 to Rothschild; 2000-02-28 to MMC/GATX and Transamerica, terminated 2003-05-09) that never moved title. No listed NPE, no defensive aggregator, and no licensing-only shell appears at any point; the technology was commercialized by the owner (PAPNET → AutoPap → SurePath/FocalPoint) and the owner litigated it against a direct competitor (Cytyc/Hologic). The two honest caveats: the Chapter 11 distress trigger at NSI is real (signal 6 present) even though it sold to an operator rather than an NPE, and signal 3 — repeat correspondent — could not be scored at all because reel/frame and correspondent data were unavailable; a manual pull at the Assignment Center could in principle change that single call, though it would not alter the ownership chain.

Verify at: https://assignmentcenter.uspto.gov/ (search patent number 5287272 — expect eight recordations: three dated 1990-01-05, one 1990-12-14, one 1999-07-26, one 2000-02-28, one 2003-05-09, one 2003-05-12) · alternate index: https://assignment.uspto.gov/patent/index.html

I'll research US patent 5287272 through the USPTO and related databases to identify its cited prior art.

Let me get the specific citation list for US5287272 from the patent databases.

Let me retrieve the actual "References Cited" section of US5287272 directly.

Prior-Art Analysis — US 5,287,272

1. The patent under analysis (as confirmed by the searched records)

Source: https://patents.google.com/patent/US5287272/en ; US5740270 patent front page (family member, same assignee/inventors).

Important scope note on the effective filing date: Because 5,287,272 is a continuation-in-part of 07/179,060 (filed Apr. 8, 1988), new matter is entitled to the Oct. 11, 1989 filing date, while matter carried over from the parent gets the Apr. 8, 1988 date. Any § 102 analysis must therefore be done claim-by-claim against whichever date the claimed subject matter is supported by.

2. Methodological caveat (please read before the tables)

I was not able to retrieve the complete, official front-page "References Cited" (56) list of US 5,287,272 itself from the sources reached in this session. The results returned the Google Patents text, the family member US 5,740,270 (a later continuation of the same chain), the EP counterpart EP 0 336 608 A3, and citing documents — but not 5,287,272's own examiner/applicant citation block.

To avoid fabrication, I have therefore separated the references into three clearly labeled provenance groups: (A) references cited in the specification text of 5,287,272; (B) references that appear in the family member US 5,740,270's "References Cited" block; and (C) references in the EPO search report for the counterpart EP 0 336 608. The date information and titles below are taken from those records. Where I am quoting a reference I could not independently inspect (e.g., the Collings and Gonzalez references, which the patent cites only by author/chapter), I say so.

3. Group A — References cited in the specification of US 5,287,272 ("Background references of interest")

These are the applicant's own cited art (non-patent literature). All pre-date the Oct. 11, 1989 filing date.

§ 102 note: None of A1–A7, standing alone, discloses the full combination claimed (primary classifier + neural-net secondary classifier + optional human tertiary review on a cytology specimen). They are best characterized as § 103/§ 102(b) art, individually or in combination, and the applicant's own "Background" discussion pre-emptively distinguishes them.

4. Group B — "References Cited" block of family member US 5,740,270 (same inventors/chain; likely overlaps 5,287,272's cited art)

The front page of US 5,740,270 (continuation filed June 7, 1995; also claiming Ser. No. 07/420,105) lists the following. Dates and numbers are reproduced literally as they appeared; I have not auto-corrected them.

U.S. Patent Documents (excerpt as retrieved):

Foreign Patent Documents: 5433555 (12/1974, Japan, G01N 33/16); 2093586 (2/1982, United Kingdom, G06K 9/80); 8909969 (10/1989, WIPO, G06F 15/18).

Other Publications (excerpt): Atkin et al., "Deoxyribonucleic Acid in Human Tumours as Measured by Microspectrophotometry of Feulgen Stain…" (1956); Caspersson, "Quantitative Cytochemical Studies on Normal, Malignant, Premalignant and Atypical Cell Populations from the Human Uterine Cervix" (1964); Ord et al., "Sample Preparation in Diagnostic Cytology" (1990); Rosenthal, "Automation of Uterine Cervical Cytology: Accomplishments and Goals" (1986).

§ 102 assessment for Group B: These are largely cell-analysis / image-cytometry hardware and measurement patents (e.g., Bacus US 4,175,860; Kopp US 4,213,036 — the Fourier-transform cell classifier discussed at length in the search results). They relate to the measurement of cell features but do not disclose the claimed neural-network secondary classifier or the primary→secondary→tertiary cascade. They are candidate § 103 art, especially for the morphological-feature-extraction elements. I cannot responsibly assert § 102 anticipation for any of them without inspecting each specification.

5. Group C — Documents cited in the EPO search report for counterpart EP 0 336 608 (priority US 179,060)

Source: EP 0 336 608 A3 European Search Report (application 89 30 2889; priority US 179,060 of 08.04.88).

Note on applicability: EP 0 336 608 corresponds to the parent application 07/179,060 (US 4,965,725), not to 5,287,272 directly. Because 5,287,272 is a CIP, the same references would be relevant only to the extent the claims of 5,287,272 are supported by the parent disclosure. Treat this group as informative about the examiner's view of the shared disclosure, not as a definitive § 102 finding against 5,287,272.

6. Bottom line on 35 U.S.C. § 102

1. No reference located in this research anticipates (fully discloses every limitation of) an independent claim of US 5,287,272. The two references the EPO examiner applied (Tien 1987 and Greenberg US 3,333,248) were both classified "Y" — i.e., evidence of obviousness in combination, not § 102 anticipation. The remaining references address isolated elements (neural nets generally, morphological image processing, cell-measurement hardware).

2. The genuinely distinguishing, claim-level elements that no single reference located supplies are: (a) the cascade of a morphological/feature-extraction primary classifier, a neural-network secondary classifier, and a human tertiary classifier, and (b) the automatic centering of the suspect nucleus in the neural net's input array via the erosion/dilation-IOD filtration (which the patent expressly identifies as the advance over the prior art). Any § 102 challenge would have to come from a single reference disclosing that combination.

3. Best § 103 candidates (not § 102): Tien et al. 1987 (C1/A2) combined with Lippmann 1987 (A4) or Hecht-Nielsen 1988 (A3) — automated cervical-smear classification + neural-net classification.

4. I could not obtain the official USPTO front-page (56) References Cited list for 5,287,272 itself. The list in Group B belongs to the sibling patent US 5,740,270, and the list in Group C belongs to the EP counterpart EP 0 336 608. If you need a certified § 102 opinion, the authoritative source is the 5,287,272 file wrapper at USPTO Patent Center (application 07/420,105) and the printed patent front page, which I was unable to retrieve verbatim in this session.

Recommended next verification steps: pull the 5,287,272 front page and file wrapper from USPTO PatentCenter (application 07/420,105) to obtain the exact examiner-cited references; and retrieve the full specifications of Greenberg US 3,333,248, Bacus US 4,175,860, and Kopp US 4,213,036 if a formal anticipation search against the primary-classifier claims is required.

I'll start by gathering the prior-art references cited on this patent and in its prosecution history, since the fetched page text doesn't include a formal "Prior Art"/claims section.

I have the claims now. Let me pull the front-page references cited and the remaining claims, plus the EP counterpart's search report (which lists the art considered against the same disclosure).

Let me retrieve the EP counterpart's search report table (same disclosure, same claims family) and check for any reexamination history.

Obviousness Analysis — U.S. Patent 5,287,272 under 35 U.S.C. § 103

0. Sourcing note (what I actually used, and what I could not verify)

The authoritative text supplied for this task (https://patents.google.com/patent/[US5287272](/patent/US5287272)/en) contains a Description, a Definitions block, IPC/CPC classifications, and a "Background references of interest" list — but no "Prior Art" tab and no claims. So I built the analysis from:

1. The four "Background references of interest" named in the patent's own Background section (Rumelhart; Tien; Hecht-Nielsen; Lippmann) plus the Collings, Gonzalez (Ch. 7) and Serra references named in the same passage.
2. The EP counterpart search report for the same disclosure (EP 0 479 977 B1) — https://patentimages.storage.googleapis.com/9c/e1/89/ba89bf4b523e99/EP0479977B1.pdf
3. The EP search report for the parent (EP 0 336 608 A3) — https://patentimages.storage.googleapis.com/c7/29/5a/7833b8bfe4c39b/EP0336608A3.pdf
4. The front-page "References Cited" list reproduced in the later family member US 5,740,270 — https://patentimages.storage.googleapis.com/2b/52/40/f56f5f367bf26a/US5740270.pdf
5. The claim text as published at https://FreePatentsOnline.com/5287272.html

Caveats I will not paper over: (a) I did not retrieve a formal PTO-1449/892 "References Cited" list specific to US 5,287,272 itself — the list at item 4 is from a continuation in the same family and may differ; (b) I did not retrieve the text of Greenberg US 3,333,248 or Clark US 4,805,225, so I describe their role only as the examiner categorized it; (c) the family record shows a B1 reexamination certificate for US 5,287,272 published 1996-08-27, but I could not retrieve its substance. I flag rather than guess.

Governing law: Application 07/420,105 was filed 1989-10-11 (CIP of Ser. No. 179,060, filed 1988-04-08), so pre-AIA § 103 applies. The critical date is 1988-04-08 for parent-supported matter and 1989-10-11 for CIP-added matter.

1. The claims at issue

Independent claims are 1 (classifier), 23 (quality-control classifier for pre-screened specimens) and 24 (method). Pertinent text:

• Claim 1: microscope means → camera means → image digitizing means → a primary classifier for identification of objects in such digital representation based on a detectable feature → secondary adaptive classifier means for recognition of cells having patterns atypical of patterns in cells expected in the specimen.
• Claim 4: secondary adaptive classifier "includes a neural network." Claim 26 is the method analogue.
• Claim 13: primary classifier "comprising means for eroding and dilating objects." Claim 21: IOD thresholding. Claim 19: primary statistical classifier. Claim 20/22: threshold/morphological criteria.
• Claims 2, 3, 25: tertiary classification and high-resolution color monitor.
• Claims 5–7: slide-information reader (barcode / OCR). Claims 8, 27: marking the location of a recognized cell.
• Claim 23: the same pipeline, but applied to specimens "which previously have been screened as not containing pre-malignant or malignant cells."

Level of ordinary skill: an image-processing/computer engineer with an automated-microscopy or cytology-image-analysis background, working with (or alongside) a cytotechnologist, ca. 1987–89.

2. Prior art available

3. Obviousness combinations a challenger would assert

Combination A — the core case against claims 1 and 24

Tien (1987) + Rumelhart (1986) / Lippmann (1987) / Hecht-Nielsen (1988).

• Tien supplies every element except "adaptive": microscope + camera + digitization + a primary screen for atypical objects + a secondary classifier + a tertiary reviewing stage, in cervical smears specifically.
• Rumelhart/Lippmann/Hecht-Nielsen supply the substitution: a trainable multi-layer backpropagation network that classifies patterns without hand-written rules, running on commercially available accelerator hardware fast enough for the volume.
• Motivation, grounded in the references themselves: EP 0 336 608 B1 records that Tien "concludes that the addition of further statistical and rules-based classifiers offers the possibility of reducing error rates," while noting this "increases complexity and therefore expense, and increases the processing time." Rumelhart/Lippmann offer exactly the tool that gets the error-rate benefit without hand-authored rules, and Hecht-Nielsen supplies the throughput that answers the processing-time objection. Under KSR v. Teleflex, 550 U.S. 398 (2007), this is a combination of known elements with a predictable result and an articulated, reference-grounded reason to try it.

Counter (stronger than it looks): the "adaptive classifier means" is not merely a bolted-on module. In the specification, the primary classifier's erosion/dilation + IOD filtration "automatically results in a centering of each suspect image around its dark centroid," so that the neural net is "presented with precisely the same type of 24 × 24 pixel images on which it was trained." The patent expressly identifies the prior-art failure mode as inability to consistently present the classifier with the centroid of the image requiring classification ("back propagation networks are excellent at reading handwritten zip code digits but have difficulty in finding where the zip code is on the envelope"). Tien supplies a sequence of stages; it does not supply this coupled centering-by-prescreen insight.

Combination B — dependents 13, 19, 20, 21, 22

Combination A + Serra (1982) + CYBEST (Tanaka 1982 / Masato 1975) + Deindoerfer (1985).

Serra supplies erosion/dilation as known morphological tools — the patent's own text concedes this ("The basic operations of erosion and dilation can be found in several sources in the prior art (e.g., Serra…)"). CYBEST and the Yellow IRIS supply automated cytological/particle prescreening by nuclear size and densitometric (integrated optical density) ranking. Motivation: eliminate false objects created by overlapping cells (dark overlap regions masquerading as dense nuclei) and reduce object count before the expensive secondary stage. This is the strongest obviousness case in the patent, because it is essentially a combination of the applicant's own acknowledged prior art.

Combination C — claims 2, 3, 25 (tertiary stage + high-res color monitor)

Combination A + Tien's tertiary stage. Tien already teaches tertiary review. Rendering stored suspect-cell images on a high-resolution color CRT is an ordinary design choice once you have stored the images. Note, however, that the patent's "field of 64 such images" review workstation is claimed in the later family members, not here — so the exposure in claims 2/3/25 is limited.

Combination D — claims 16, 17, 18 (automated microscope, video camera, CCD)

Combination A + US 4,700,298 (Palcic) and/or US 4,805,225 (Clark). Both are automated-microscope/image-scanning systems. These are peripheral, largely conventional hardware recitations.

Combination E — claims 5, 6, 7 (barcode / OCR reader)

Combination A + a conventional identification reader. The patent itself concedes the barcode reader is "a conventional device, such as… sold by Symbol Technologies, Inc… under the trademark LaserScan 6X20," and expressly contemplates "an optical character reader" as an equivalent. The only genuine question is motivation: reducing transcription error and preventing specimen/report mismatch. I could not locate, in the prior-art material available to me, a reference directed to slide-borne machine-readable identification in an automated cytology pipeline. A challenger would have to rely on general knowledge of laboratory information systems; that is a comparatively weak showing.

Combination F — claims 8, 27 (marking cell location)

Combination A + Greenberg US 3,333,248, as the EP examiner paired Tien and Greenberg against the parent's claims 1–3, 8–11. The patent itself analogizes to semiconductor-wafer marking. I have not verified Greenberg's disclosure, so I do not assert that it teaches slide marking.

Combination G — claim 23 (QC/rescreen of previously-screened-negative slides)

No adequate combination is available from the identified art. Tien, CYBEST and the Yellow IRIS all presume primary screening. Claim 23's preamble — specimens "which previously have been screened as not containing pre-malignant or malignant cells" — requires a specific use context none of the listed references discloses. A challenger would have to argue it is obvious to run an automated screen on slides a human already called negative, motivated by the known false-negative rate. That is a plausible "obvious to try" argument under KSR, but it is the weakest combination in this set and the one I would expect to fail.

4. Why the whole may nonetheless be non-obvious

1. Teaching away / field skepticism. The Background recites two decades of failure and frames the field's accepted workaround as monolayer preparation to avoid overlapping cells. The listed art (CYBEST, Yellow IRIS, GB 2 093 586, Bacus, Palcic) all operate on comparatively clean, non-overlapping or single-cell populations. Reading them as a whole points away from attacking standard Pap smears.
2. The prior art's own arithmetic cuts against the combination. Hecht-Nielsen/Lippmann frame neural nets as computationally intensive; Tien warns added classifiers increase processing time. A POSITA reading these together would more likely conclude that a neural net secondary stage cannot fit the time budget — which is precisely the obstacle the patent says it overcame.
3. The non-obvious element is the cou[pling], not the parts. Combination A reconstructs the architecture from the outside but not the automated prescreen-centering that makes the adaptive classifier work. That is the point of novelty the specification repeatedly claims.
4. Secondary considerations. The Background asserts unexpected success on standard Pap smears with overlapping/partially obscured cells, operation within manual-screening time (~6 min/specimen as recited), and lower false-negative rates than manual screening. The family's stock of later patents (US 5,544,650; US 5,740,270; US 5,939,278; US 6,327,377) and the 1996 B1 reexamination certificate for this patent are consistent with commercial significance and with the claims surviving an Office validity challenge — though I did not retrieve the certificate and cannot state what it did.
5. § 103(c) caveat for US 4,965,725. The parent was filed 1988-04-08 and issued 1990-10-23, i.e., after this application's 1989-10-11 filing date, so it could only be § 102(e)-type art. Pre-AIA § 103(c) bars use of § 102(e)/(f)/(g) art for obviousness where the subject matter was commonly owned at the time of invention; both were assigned to Neuromedical Systems, Inc. (assignments of record 1990-01-05). A challenger would have to work around § 103(c), and the same-family relationship also raises obviousness-type double patenting rather than § 103.

5. Bottom line

• Independent claims 1 and 24 (method): a prima facie § 103 case is constructible from Tien + Rumelhart/Lippmann/Hecht-Nielsen, and it is a legitimate case under KSR. But it requires the challenger to treat the neural-net stage as a black box and ignore the prescreen-centering function the specification identifies as the critical advance. I assess this as contestable but not clearly obvious — I would expect the patentee to prevail on the coupling argument plus secondary considerations.
• Dependent claims 13, 19, 20, 21, 22 (erosion/dilation; IOD threshold; morphological criteria): these are the most exposed, because they recite techniques the patent itself cites as known (Serra; CYBEST; Yellow IRIS). If claims 1/24 are found obvious, these offer no independent vitality.
• Claims 5–7 (barcode/OCR), 8/27 (marking), 16–18 (microscope/camera/CCD): weak prima facie cases unless a specific secondary reference is supplied. Claim 6's own admission that the barcode reader is an off-the-shelf device makes claim 5 obvious only if a motivation reference exists — which I could not find in the available art.
• Claim 23 (QC rescreen of pre-screened slides): no viable § 103 combination on the art identified here.
• Recommended framing for a challenge: lead with Tien + Lippmann/Rumelhart on claim 1; add Serra + Tanaka (CYBEST) + Deindoerfer for claims 13 and 21; add Greenberg US 3,333,248 for claims 8/27 (subject to verifying its disclosure). Do not rely on US 4,965,725 as § 103 art — § 103(c) common ownership is a serious obstacle, and I would need to confirm ownership as of the invention date before asserting it.

Confidence: high on the claim text and on the EP search-report citations (both retrieved and quoted verbatim); moderate on the front-page reference list (it comes from US 5,740,270, a different family member); low/flagged on Greenberg US 3,333,248, Clark US 4,805,225, and the 1996 reexamination certificate, none of whose texts I retrieved. I have not asserted anything about those three beyond how the record categorizes them.