Dental Monitoring SAS v. Align Technology — Federal Circuit Holds AI-Powered Dental Imaging Patents Ineligible Under § 101

Background

Dental Monitoring SAS, a French company focused on remote orthodontic care, owned two patents — U.S. Patent No. 11,049,248 (‘the ‘248 patent’) and U.S. Patent No. 10,755,409 (‘the ‘409 patent’) — covering methods for analyzing dental arch images using a “deep learning device.” In plain terms, the patents described using a neural network, trained on thousands of labeled dental images, to assess how well an orthodontic aligner (such as an Invisalign tray) fits a patient’s teeth and to guide clinicians in capturing usable images.

In November 2022, Dental Monitoring sued Align Technology — the maker of Invisalign — in the Northern District of California, alleging that Align’s Invisalign Virtual Care AI platform infringed both patents. Judge William Alsup organized the case as a “patent showdown”: each side selected a representative claim, underwent targeted discovery, and then filed cross-motions for summary judgment. Dental Monitoring chose claim 14 of the ‘248 patent; Align chose claim 12 of the ‘409 patent. The parties agreed the court’s rulings on those “showdown claims” would bind the remaining asserted claims.

In May 2024, Judge Alsup granted summary judgment for Align, concluding both patents were directed to patent-ineligible abstract ideas under 35 U.S.C. § 101. Dental Monitoring appealed. The Federal Circuit — in a nonprecedential opinion authored by Circuit Judge Alan Lourie, joined by Judges Schall and Taranto — affirmed.

The Court’s Holding

The Federal Circuit applied the familiar two-step Alice framework (Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (2014)) to find both patents invalid.

Step One — directed to an abstract idea: Claim 14 of the ‘248 patent is directed to “collecting and analyzing image information using a deep learning device” — a process orthodontists have always performed, now simply delegated to a neural network. Claim 12 of the ‘409 patent is similarly directed to the abstract idea of “image acquisition, analysis via a ‘deep learning device,’ comparison to a setpoint, and transmittal of the analysis result.” Both fall squarely within the familiar category of ineligible claims that “gather and analyze information of a specified content, then display the results.”

The court emphasized — citing its 2025 decision in Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025) — that deploying a generic machine learning model in a new domain (here, dental arch imaging) does not make the claims patent-eligible. Training a neural network on 1,000-plus dental images is “incident to the very nature of machine learning,” not a technological improvement. And the patents themselves confirmed the generic character of the “deep learning device” by pointing to off-the-shelf neural networks from Google and Microsoft as suitable implementations.

Step Two — no inventive concept: Because the “deep learning device” is conventional, it adds nothing “significantly more” beyond the abstract idea. Dental Monitoring’s argument that using machine learning in orthodontics was “unconventional” at the time of issuance was rejected: the eligibility inquiry asks not whether the whole invention is new, but whether the claim elements add an inventive concept that transforms the abstract idea. They do not.

Key Takeaways

• New field ≠ patent-eligible AI claim. Applying an existing machine learning architecture to a new industry vertical — dental imaging, medical diagnostics, financial modeling — will not satisfy § 101 under Recentive Analytics.
• Domain-specific training data is not an inventive concept. Claiming that a neural network must be trained on thousands of domain-specific images (dental arches, histology slides, satellite photos) does not escape the abstract-idea bar; such training is inherent to how machine learning works.
• Own specification can sink the case. Dental Monitoring’s patents listed Google and Microsoft neural networks as suitable “deep learning devices,” confirming their generic nature — and the Federal Circuit used those very disclosures against the patent holder.
• Efficiency gains are not enough. The court reaffirmed that performing a task “with greater speed and efficiency” through machine learning does not confer patent eligibility.

Why It Matters

This decision adds another data point to the Federal Circuit’s post-Recentive Analytics body of law on AI patent eligibility. Companies commercializing AI-powered tools in healthcare, imaging, or other regulated industries should note that claiming a “deep learning device” applied to a new domain — without a specific algorithmic or technical advance beyond the abstract concept of “collect, analyze, display” — is an uphill battle under § 101. Patent prosecutors should draft claims that emphasize concrete technical improvements in the AI system itself, not just the field of application.

Although nonprecedential, this opinion is one of three related Dental Monitoring-vs.-Align cases currently pending before the Federal Circuit (see also Nos. 2025-1752 and 2025-1879), and it signals how the court is likely to approach similar AI-patent eligibility questions as that docket continues.