Federal Circuit Finds Deep Learning Patents Ineligible

The U.S. Court of Appeals for the Federal Circuit (CAFC) issued a decision today in Dental Monitoring SAS v. Align Technology, Inc., affirming a district court ruling that found several patent claims covering deep learning based dental image analysis invalid as directed to ineligible subject matter under Section 101.

Dental Monitoring SAS owns U.S. Patent 11,049,248 and U.S. Patent 10,755,409, both of which relate to dental arch image analysis. The ‘248 patent covers “a method for assessing the shape of an orthodontic aligner using a deep learning device,” while the ‘409 patent covers a method for acquiring and analyzing an image of a dental arch using the same type of device. In November 2022, Dental Monitoring filed a lawsuit against Align Technology, Inc. in the U.S. District Court for the Northern District of California before Judge William Alsup, alleging that Align’s Invisalign Virtual Care AI platform and related apparatuses infringed the ‘248 and ‘409 patents, as well as a third patent not at issue on appeal.

In July 2023, the district court structured the case as what it called a “patent showdown,” directing each side to choose one claim for cross-motions for summary judgment following discovery. Dental Monitoring selected claim 14 of the ‘248 patent and Align selected claim 12 of the ‘409 patent. They both stipulated that the ruling on those claims would also apply to claim 1 of the ‘248 patent and claims 1 and 7 of the ‘409 patent. In January 2024, Dental Monitoring argued that Align infringed the selected claims as a matter of law. Align responded that the claims were ineligible under Section 101 and, alternatively, invalid under 35 U.S.C. Section 112.

Applying the two-step framework from Alice Corp. v. CLS Bank International, the district court sided with Align. At step one, the district court found that claim 14 of the ‘248 patent was directed to “collecting information, analyzing it, and displaying certain results of the collection and analysis,” quoting the CAFC’s earlier decision in Electric Power Group, LLC v. Alstom S.A. Regarding the ‘409 patent, the district court found that independent claim 1 was directed to the abstract idea of acquiring and analyzing an image, then generating a message to guide a user in taking a new image, reasoning that the claim “merely recite[d] a common practice long performed by dental practitioners.”

At step two, the district court found no inventive concept in either patent, concluding that the claims used generic hardware to accomplish the abstract idea within the specific field of dental aligner assessment. The district court granted summary judgment of ineligibility, and Dental Monitoring appealed.

Writing for the court, Judge Lourie, at Alice step one, agreed that both sets of claims were directed to abstract ideas. Claim 14 of the ‘248 patent was found to be directed to the abstract idea of collecting and analyzing image information through a deep learning device, while claim 12 of the ‘409 patent was found to be directed to acquiring an image, analyzing it through such a device, comparing the result to a setpoint, and transmitting the outcome. The court placed both within the “‘familiar class of [patent-ineligible] claims’ that focus on ‘collecting information, analyzing it, and displaying certain results of the collection and analysis,” again citing Electric Power Group.

Dental Monitoring argued that claim 14 of the ‘248 patent reflected a specific technological solution because the trained device could quantitatively assess the separation between an aligner and a tooth with more precision than previously possible. The CAFC rejected that argument, explaining that the claim language itself did not require any particular degree of quantitative precision beyond what an orthodontist could already achieve without a deep learning device. The court also found that training the device on a dataset of more than 1,000 images did not amount to a technological improvement, since “a ‘deep learning device’ be trained on a specific subset of data is ‘incident to the very nature of machine learning.” That reasoning drew on the CAFC’s decision in Recentive Analytics, Inc. v. Fox Corp., which held that applying generic machine learning tools to a new field of use, even one previously performed by humans, does not confer eligibility.

At Alice step two, the CAFC found no inventive concept sufficient to save the claims. The specifications of both patents confirmed that the deep learning device could be selected from a preset list of widely available neural networks, including networks associated with major technology companies. Since the claims applied that generic device to carry out the same abstract idea identified at step one, the court found nothing that transformed the claims into something “significantly more” than a claim on the ineligible concept itself, citing Broadband iTV, Inc. v. Amazon.com, Inc.

The CAFC also rejected Dental Monitoring’s argument that using deep learning devices for orthodontic guidance was not conventional when the patents were issued. The court, citing BSG Tech LLC v. BuySeasons, Inc., explained that the relevant inquiry at step two is whether the claims contain an inventive concept, not whether the invention as a whole was unconventional. It found no genuine factual dispute precluding summary judgment, since the patents’ own specifications confirmed the generic nature of the deep learning device.

Ultimately, the CAFC affirmed the district court’s grant of summary judgment, holding that claims 1 and 14 of the ‘248 patent and claims 1, 7, and 12 of the ‘409 patent are ineligible under Section 101 and therefore invalid. Since the court agreed with Align on eligibility, it did not reach Align’s alternative Section 112 argument.

Image Source: Deposit Photos
Author: alexmillos
Image ID: 35440567