All Bark, No Bite: Federal Circuit Rejects Dental-Related Machine Learning Claims as Too Generic

The decision highlights the importance of documenting technical innovations during product development and how having technical and legal teams collaborate early in the process can strengthen patent protection for AI-enabled medical devices.

In a nonprecedential opinion, the US Court of Appeals for the Federal Circuit affirmed a district court’s finding that claims 1 and 14 of US Patent No. 11,049,248 and claims 1, 7, and 12 of US Patent No. 10,755,409 were ineligible under 35 USC § 101.[1] The Federal Circuit did not announce a categorical rule against patenting deep learning inventions. Instead, it concluded that the asserted claims recited a generic deep learning device that performed abstract information-processing functions without reciting a specific technological improvement. The decision illustrates how the Federal Circuit applies existing eligibility doctrine to AI-related claims, particularly in digital health and diagnostics contexts.

Key Takeaways

• The Federal Circuit affirmed a lower court’s finding that asserted machine learning claims were ineligible because they were directed to an abstract idea and the claims did not recite sufficient details regarding the “deep learning devices” to supply an inventive concept.
• The decision does not bar patents covering AI-enabled medical devices. However, it highlights that medical device companies should connect legal, technical, clinical, regulatory, and quality teams early in the product development process to ensure that patents claim and describe the device’s inventive concept.

OVERVIEW

Dental Monitoring SAS sued Align Technology, Inc. in November 2022, alleging that Align’s Invisalign Virtual Care AI platform infringed claims of the ’248 and ’409 patents.[2] The ’248 patent claimed a method for evaluating the shape of a patient’s orthodontic aligner using a “deep learning device.”[3] The ’409 patent claimed a method for acquiring an image of a dental arch, analyzing the image using a “deep learning device,” and determining the degree of separation between the aligner and the patient’s teeth.[4]

In July 2023, the district court required each party to select one claim, conduct focused discovery, and then file cross-motions for summary judgment.[5] Dental Monitoring selected claim 14 of the ’248 patent and Align selected claim 12 of the ’409 patent.[6] The parties stipulated that the rulings on those claims would also govern the other asserted claims. In May 2024, the district court granted Align’s motion for summary judgment of patent-ineligibility under 35 USC § 101.[7] Dental Monitoring appealed.[8]

Analysis of the Federal Circuit’s Decision

In Alice Corp. v. CLS Bank International, the Supreme Court set forth a two-step framework for evaluating patent eligibility under § 101, which defines the subject matter eligible for protection as any “new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof.”[9] First, a court asks whether a claim is directed to a patent-ineligible concept, such as an abstract idea.[10] If so, the court then asks whether the claim adds an “inventive concept” that transforms the abstract idea into a patent-eligible application.[11] The Federal Circuit applied the familiar Alice standard to determine that the asserted claims were not patent eligible under § 101.[12]  

Step One: Claims Recited the Abstract Idea of Image Collection and Analysis Using a ‘Deep Learning Device’

At step one, the Federal Circuit held that the claims fell within “the ‘familiar class of patent-ineligible claims’ that focus on ‘collecting information, analyzing it, and displaying certain results[.]’”[13] That the claims recited a “deep learning device” for analyzing images did not change the court’s analysis because the patents’ specifications generically described the “deep learning device.”[14] Nor was it sufficient that the claims required that the “deep learning device” be trained in a specific manner (i.e., on images of dental arches) because training such devices on data subsets, according to the Federal Circuit, is “‘incident to the very nature of machine learning,’” not a “‘technological improvement.’”[15]

Dental Monitoring argued that claim 14 provided a “specific technological solution” because the trained device assessed aligner fit on a “tooth-by-tooth basis” and did so more precisely, objectively, and efficiently than was previously possible.[16] The Federal Circuit rejected this argument on two grounds. First, the court noted that the claim neither required a particular degree of quantitative precision nor improved how the deep learning device performed the analysis.[17] The court then noted that the asserted claims were not rendered patent eligible merely because they “‘perform a task previously undertaken by humans with greater speed and efficiency than could previously be achieved.’”[18]

Step Two: Generic ‘Deep Learning Device’ Did Not Provide an Inventive Concept

At step two, the Federal Circuit held that there was no “‘inventive concept’ sufficient to transform the ineligible abstract idea into patent-eligible subject matter” because the patents described the “deep learning device” as selectable from “a list of well-known and available neural networks.”[19] This supported the conclusion that the device was generic and undercut Dental Monitoring’s argument that there was a factual dispute about the generic nature of the claimed device that precluded summary judgment.[20]

Dental Monitoring argued that the claims recited an inventive concept because the “deep learning device” was trained in a specific process.[21] The Federal Circuit rejected this argument, explaining that the device’s use of specific training images did not make the “‘deep learning device’ itself non-generic.”[22]

Dental Monitoring also argued that using “deep learning devices” in orthodontic treatment was unconventional when the patents issued.[23] The Federal Circuit disagreed, noting that the inquiry under Alice step two is not whether the claimed invention as a whole was novel or non-routine, but whether the claims recite an inventive concept that transforms the abstract idea into “something ‘significantly more’ than the abstract idea.’”[24]

Practical Implications

AI-Based Claims Require a Degree of Specificity Regarding the Innovation

The Federal Circuit’s decision highlights the importance of claiming details regarding the technological innovation underlying AI-based inventions, including machine learning algorithms. AI-based claims are regularly subject to § 101 challenges in part because they often involve steps for collecting and analyzing data. In this instance, the claims were challenged under § 101 even though they recited a “deep learning device” that specialized in image classification and required the device to “be trained in a specific manner” using “more than a thousand images of dental arches.”[25] Notably, the Federal Circuit found both of these factors too “generic” and thus insufficient to transform the claims into patent-eligible inventions. The decision highlights the need for applicants and practitioners to ensure that claims contain sufficient detail regarding the novel approach to using a machine learning model to withstand scrutiny, especially if the model itself provides the technological innovation.

Documenting the Innovation Is Key

The Federal Circuit’s holding also highlights the importance of carefully documenting technical innovations during product development—not just during patent prosecution or litigation. By contemporaneously documenting the specific technical problems they encounter, the solutions they implement, and the resulting improvements, engineering teams can help ensure that their innovations make their way into patent applications and are well-positioned to withstand eligibility challenges. This documentation should not merely describe how efficiently or accurately an AI model performs. Instead, it should describe the technical mechanism that produces the claimed improvement (e.g., novel approaches to system architecture, data acquisition, or image processing).

Involve Patent Counsel Early

For AI-enabled medical devices, patent strategy should begin during the product development phase. Engineering, clinical, quality, regulatory, and patent teams should collaborate to identify specific technical advances over prior systems and to document those advances as they are developed. This will enable patent counsel to identify technical improvements that support patent eligibility, ensure that those improvements are adequately described in the specification, and develop claims directed to those improvements rather than to the use of a generic AI model or machine learning device.

Conclusion

The Dental Monitoring decision reinforces that claims generically reciting AI or machine learning devices are unlikely to withstand eligibility challenges under § 101. Companies developing AI-enabled medical and digital health technologies should identify, document, and describe the underlying technological innovation—not merely the use of AI to achieve a better result. By involving patent counsel early and carefully documenting innovations, applicants can place themselves in a stronger position to defend their patents many years later.

How we can help

Our AI and medical device teams help legal, R&D, and regulatory leaders identify protectable technical advances early and develop patent applications that document the technical reason a product performs better. We also review existing portfolios for eligibility and enforcement risk, evaluate follow-on filing strategies, and align patent protection with product development, licensing, transactions, and litigation objectives.