Ipr In AI-Assisted Diagnostic Device Patents.
1. AI-Assisted Diagnostic Devices and IPR
AI-assisted diagnostic devices combine artificial intelligence algorithms with medical devices or software to help detect, diagnose, or predict medical conditions. Patenting such devices involves addressing:
Patent Eligibility
Whether AI algorithms are considered patentable subject matter.
Intersection of medical methods (often excluded in many jurisdictions) and AI processes.
Inventive Step / Non-obviousness
The AI model’s novelty and non-obvious technical contribution.
Differentiation from standard clinical practices or generic computing steps.
Disclosure Requirements
The patent must explain the AI model clearly (e.g., architecture, training data, and application) so that someone skilled in the art can replicate it.
2. Key Legal Issues in AI Diagnostic Patents
Software Exclusion: Many jurisdictions, like the U.S. (35 U.S.C. §101) and India, exclude pure software or abstract ideas from patents. AI algorithms must show technical effect or integration with a medical device.
Medical Method Exclusion: Pure medical procedures may not be patentable, but AI-assisted methods tied to a device or apparatus often are.
Obviousness: Courts often evaluate whether the AI solution is merely a routine application of known algorithms or offers a true inventive step.
Data and Training Set Protection: IP does not automatically cover the training data; it protects the algorithmic approach and model output.
3. Important Case Laws
Case 1: Mayo Collaborative Services v. Prometheus Laboratories, Inc., 566 U.S. 66 (2012) (USA)
Facts: Prometheus patented a method of optimizing drug dosages based on metabolite levels. Mayo argued it was a law of nature.
Ruling: The U.S. Supreme Court held that simply applying a natural law using standard steps is not patentable.
Relevance to AI Diagnostics:
AI algorithms predicting disease outcomes from biological markers may be treated as “abstract ideas” or “laws of nature.”
Patents must show an inventive application of AI, not just correlating data with a diagnosis.
Case 2: Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014) (USA)
Facts: Alice Corp. patented a computer-implemented scheme for mitigating financial risk.
Ruling: Court introduced the Alice test for software patents: abstract idea + inventive concept.
Relevance:
For AI diagnostic devices, the algorithm itself is often abstract.
A successful patent must integrate the AI into a technical application, e.g., a novel imaging device for tumor detection.
Case 3: Enfish, LLC v. Microsoft Corp., 822 F.3d 1327 (Fed. Cir. 2016)
Facts: Enfish patented a self-referential database.
Ruling: The court ruled that a software invention is patentable if it improves computer functionality, not just automates a process.
Relevance:
AI-assisted diagnostic devices can be patentable if they improve the functionality of diagnostic machines, e.g., faster detection or higher accuracy than prior methods.
Case 4: Thales Visionix, Inc. v. United States, 850 F.3d 1343 (Fed. Cir. 2017)
Facts: Patent on sensor fusion in motion-tracking systems.
Ruling: Patents for integrating sensors and software into devices were upheld because they solved a technical problem.
Relevance:
AI diagnostic devices integrating imaging, sensor data, and predictive algorithms can qualify as patentable technical solutions.
Case 5: Mayo and AI-Driven Diagnostics (Follow-up)
Some AI-based diagnostic patent applications were rejected in the USPTO citing Mayo/Alice precedent:
Example: AI model for predicting cardiovascular disease risk using patient data.
USPTO held: merely training an AI model on medical data is abstract; patentable only if applied in a novel technical method/device.
Case 6: Indian Context – Novartis AG v. Union of India (2013) 6 SCC 1
Facts: Patentability of pharmaceutical inventions with known compounds.
Ruling: India rejected patents for minor modifications lacking inventive step.
Relevance:
AI-assisted diagnostic devices in India must show significant inventive step beyond conventional diagnosis.
AI algorithm plus device integration can be patentable if it improves accuracy, speed, or reduces invasive procedures.
Case 7: EP 2771468 – AI Diagnostic Imaging Patent (European Patent Office)
Facts: Patent for AI model enhancing MRI images to detect lesions.
Outcome: EPO granted patent emphasizing technical effect of the AI (improving image quality), not just the algorithm.
Relevance:
Shows how AI applied to medical imaging can be patentable in Europe.
Technical contribution is key: improving a device’s performance rather than just processing data.
4. Summary of Patentability Criteria for AI Diagnostics
| Jurisdiction | Key Requirement | Example/Case |
|---|---|---|
| USA | Technical application + inventive concept | Enfish, Alice |
| Europe | Technical effect / improvement to device | EP 2771468 |
| India | Inventive step + novel application | Novartis v. India |
| General | Abstract ideas, natural laws, or medical methods alone are not patentable | Mayo v. Prometheus |
5. Key Takeaways
AI algorithm alone ≠ patentable; must be tied to a technical application.
Data correlation or prediction ≠ inventive step; improvement of device performance is critical.
Disclosure is essential: patent must describe model, data, and device integration clearly.
Cross-jurisdiction differences: EPO focuses on technical effect; USPTO focuses on abstract idea + inventive concept; India emphasizes inventive step.
RELATED Blog
comments