The recent decision in Dental Monitoring SAS v. Align Technology, Inc. is an important reminder for digital health companies, developers of software as a medical device, and companies using artificial intelligence in medical products: the mere use of artificial intelligence or deep learning is not enough to establish patent eligibility in the United States.
The decision forms part of a broader trend in the case law of the United States Court of Appeals for the Federal Circuit, particularly following Recentive Analytics v. Fox Corp. Under this line of authority, a patent will not be granted merely for applying known machine-learning techniques to a new field of use, unless the claimed invention includes a concrete technological improvement in the model, architecture, training process, or computing system.
This message is especially important in the healthcare sector. Many AI-based medical systems follow a similar pattern: collecting medical information, analysing it using a model, and generating an output such as a diagnosis, alert, recommendation, risk score, or assessment of treatment progress. Following Dental Monitoring, patent claims may be vulnerable if they primarily describe the clinical or commercial result, rather than the specific technological manner in which the system achieves that result.
Dental Monitoring sued Align Technology, the maker of Invisalign, alleging that Align’s Virtual Care AI system infringed its patents relating to the remote monitoring of orthodontic treatment through the analysis of dental images.
The patents at issue on appeal concerned, among other things, a method for assessing the shape of an orthodontic aligner and a method for acquiring an image of a dental arch and analysing it using a deep-learning model. Such systems are intended to enable remote assessment of treatment progress and reduce the need for frequent in-person visits to the dentist or orthodontist.
The United States District Court for the Northern District of California granted summary judgment, holding that the relevant claims were patent-ineligible under Section 101 because they were directed to an abstract idea and did not include a sufficient inventive concept. The Federal Circuit affirmed.
Section 101 of the U.S. Patent Act concerns the threshold question of patent eligibility, namely whether the claimed subject matter is, in principle, the type of invention that can be protected by a patent. Under this provision, the court examines whether the invention falls within one of the eligible statutory categories, such as a process, machine, manufacture, or composition of matter, or whether it falls within one of the judicially created exceptions, such as an abstract idea, law of nature, or natural phenomenon.
In the context of software and artificial intelligence, Section 101 often operates as an early filter against patent claims that primarily describe the collection of information, its analysis, and the generation of a result, without a concrete technological contribution.
In Europe, the comparable question is examined mainly under Article 52 of the European Patent Convention, which excludes, among other things, computer programs “as such”, while allowing protection for computer-implemented inventions that make a technical contribution.
In Israel, the starting point is Section 3 of the Patents Law, which requires the invention to be in a “technological field”, together with the practice of the Israel Patent Office regarding computer-implemented inventions, software, and artificial intelligence. Although the legal tests differ from jurisdiction to jurisdiction, the common thread is the need to show more than a general idea or mere information processing, and to identify a genuine technological solution.
The Federal Circuit applied the familiar Alice/Mayo framework. At the first step, the court considered whether the claims were directed to an abstract idea. It answered that question in the affirmative. With respect to one patent, the court held that the claim was directed to collecting and analysing information from an image using a deep-learning model. With respect to the other patent, the court held that the claim was directed to acquiring an image, analysing it using deep learning, comparing it to a target value, and transmitting the result of the analysis.
The court treated the patent claims as falling within the well-established category of ineligible claims directed to collecting information, analysing it, and displaying certain results. The fact that the information was dental information, and that the analysis was performed using deep learning, did not change the abstract character of the claims.
At the second step, the court considered whether the claims included an inventive concept sufficient to transform the abstract idea into a patent-eligible application. Again, the answer was no. Although the claims referred to a deep-learning device trained on a learning set of more than one thousand dental images, the court held that training a model on a dataset is inherent in the nature of machine learning. Accordingly, training the model on dental images did not, by itself, constitute a specific technological solution.
The decision continues the approach taken in Recentive Analytics v. Fox Corp., where the Federal Circuit held that patents applying generic machine-learning techniques to event scheduling and network mapping were not eligible under Section 101. In that case, the patents were found to be directed to the abstract idea of using a generic machine-learning technique in a particular environment, without a sufficient inventive concept.
Recentive was significant because it was one of the first Federal Circuit decisions to address directly the patent eligibility of machine-learning inventions. Dental Monitoring applies the same reasoning in a medical-dental context: even where the field of application is healthcare, and even where there may be clear clinical or operational value, it is not enough to apply generic AI to a new category of data.
The healthcare sector is particularly exposed to this trend. Many companies are developing AI systems for medical image analysis, digital pathology, remote monitoring, patient triage, clinical decision support, early detection, treatment management, dose optimisation, anomaly detection, and software as a medical device.
In many cases, the commercial value of the system lies in its ability to generate clinical insight from medical data, such as an image, physiological signal, medical record, laboratory data, or information collected from a connected device. However, for purposes of Section 101, the mere transformation of raw medical information into clinical insight using an AI model may be viewed by a court as abstract information analysis, unless the claim explains the concrete technological improvement.
A claim framed as receiving a medical image, analysing it using a neural network, and generating a diagnosis or recommendation may therefore be vulnerable. By contrast, a claim that describes a specific technological mechanism for improving image quality, reducing noise, addressing biased or incomplete datasets, reducing false positives, improving response times, enabling more efficient training, or technically integrating a medical device, server, and clinical interface may be in a stronger position.
It is important not to read Dental Monitoring too broadly. The decision does not hold that all AI inventions in healthcare are patent-ineligible. Nor does it exclude patents on deep learning, computer vision, or software as a medical device as a matter of principle.
The message is more precise: artificial intelligence can form part of a patent-eligible invention, but the applicant must show that the invention is not merely the use of a known tool to analyse information in a new medical field. A clear technological story is required: what was the technological problem, what is the technological solution, and how does the patent reflect that solution?
This approach is also consistent with the general direction of the U.S. Patent and Trademark Office’s guidance on the eligibility of AI-related inventions, which emphasises the need to examine whether the invention is integrated into a practical application and whether the claim limitations add more than a general instruction to apply an abstract idea using a computer or artificial intelligence.
Dental Monitoring v. Align is an important signal for healthcare companies using artificial intelligence. The court was not persuaded by the mere use of deep learning, even in a commercially important medical field. To protect AI inventions in healthcare, patent applicants should frame them as concrete technological solutions, not merely as the application of machine learning to medical information.
The lesson for patent applicants is clear: the closer a patent comes to a general description of collecting information, analysing it, and producing a result, the greater the Section 101 risk. The more the patent describes a detailed, measurable, and non-generic technological mechanism, the better its chances of surviving the patent eligibility analysis.
