ARTIFICIAL INTELLIGENCE IN HEALTH CARE: APPLICATIONS AND LEGAL ISSUES

• Published in: The SciTech Lawyer Vol. 14; no. 1; pp. 10 - 13
• Main Author: Price, Nicholson
• Format: Newsletter Journal Article
• Language: English
• Published: Chicago American Bar Association 22.09.2017
• Subjects: Algorithms; Analysis; Artificial intelligence; Attorneys; Autism; Big Data; Bioethics; Clinical trials; Federal regulation; Health aspects; Health care policy; Innovations; Intellectual property; Laws, regulations and rules; Legal medicine; Machine learning; Medical equipment; Medical protocols; Medical technology; Medicine; Patients; Privacy; Skin cancer; Smartphones; Software; Torts; X-rays
• ISSN: 1550-2090

Machine learning is most familiar in the context of image recognition, and an algorithm has already been developed that can identify skin cancer by analyzing images of skin lesions; the algorithm performs as well as board-certified dermatologists.4 A recent New England Journal of Medicine article suggests that such algorithms could soon enter widespread use in image analysis, aiding or displacing much of the work of anatomical pathologists or radiologists within the span of years.5 Another current algorithm can predict which trauma victims are likely to hemorrhage by constantly analyzing vital signs and can in turn call for intervention to forestall catastrophe; such prognostic algorithms could come into use in a similarly short time frame.6 A bit farther off, black-box algorithms could be used for diagnosis more generally, to recommend off-label uses for existing drugs, to allocate scarce resources to patients most likely to benefit from them, to detect fraud or problematic medical behavior, or to guide research into new diseases or conditions. [...]blackbox algorithms are already in use today in smartphone apps that aim to identify developmental disorders in infants based on facial features7 or autism in young children based on eye movement tracking.8 The potential for benefit from such blackbox medicine is substantial, but it comes with its own challenges: scientific and medical, certainly, but also legal. [...]in black-box medicine, traditional methods of testing new medical technologies and devices are likely not to work at all in some instances, and to slow or stifle innovation in others. A string of recent decisions by the U.S. Supreme Court interpreting section 101 of the Patent Act, which governs patentable subject matter, has made it very difficult to patent blackbox algorithms.18 In Mayo Collaborative Services v. Prometheus Laboratories, Inc., the Supreme Court repeated its longstanding statement that laws of nature cannot be patented.19 However, the Court applied that rule to a diagnostic test that used the measurement of a metabolite level in a patient's blood to adjust the dosage of a drug, which many, including the Federal Circuit below, had thought to be a patentable application of such a law.