Is the Clinical Implementation of In-House Artificial Intelligence-Developed Algorithms Happening?

David Gergely Kovacs, Claes Nøhr Ladefoged, Jacob Bak Rosenkjær, Fatima Mawassi, Line Akiti Hingelberg, Anna Olga Aaskilde Laursen, Barbara Malene Fischer, Flemming Littrup Andersen

Research output: Contribution to journalJournal articleResearchpeer-review

3 Downloads (Pure)

Abstract

The advent of artificial intelligence (AI) in medical imaging and nuclear medicine could revolutionize our approach to diagnosing and treating complex conditions. Our recent paper, “Clinical Evaluation of Deep Learning for Tumor Delineation on 18F-FDG PET/CT of Head and Neck Cancer,” (1) showcases the potential of in-house AI in clinical practice. Clinically, we often find ourselves in need of a legitimate path to apply in-house developments. Here, Article 5(5) of the European Medical Device Regulation (MDR) 2017/745 regulations (2), the “in-house rule,” may provide a solution. The in-house rule allows health institutions to develop and use devices solely within their facilities, reducing the need for exhaustive external documentation and oversight typical of full conformité européenne marking. By focusing on a tailored quality management system (QMS) for safety and performance, institutions can save considerable time and resources compared with the comprehensive requirements for commercial distribution.

In line with Saboury et al.’s call for trustworthy AI frameworks (3), this article elucidates how the in-house rule supports structured and reliable clinical implementation and testing of our in-house AI algorithm. Please note that this article is intended for inspiration and should not be used as a direct guide or precedent for applying the in-house rule in practice. Interpretation and implementation will vary by institution and country because of national regulatory nuances. By sharing our experience as one of the first adopters in Denmark’s Capital Region, we hope to inspire and encourage other jurisdictions to adopt approaches that support hospital-based innovation within the framework of Article 5(5).
Original languageEnglish
JournalJournal of Nuclear Medicine
Number of pages3
ISSN0161-5505
DOIs
Publication statusAccepted/In press - 2025

Fingerprint

Dive into the research topics of 'Is the Clinical Implementation of In-House Artificial Intelligence-Developed Algorithms Happening?'. Together they form a unique fingerprint.

Cite this