Standardisation and harmonisation efforts in quantitative imaging: Commentary

Research output: Contribution to journalResearch articleContributedpeer-review

Contributors

  • Alex Zwanenburg - , OncoRay - National Centre for Radiation Research in Oncology, German Cancer Research Center (DKFZ), National Center for Tumor Diseases (NCT) Dresden (Author)

Abstract

Peel 500 grams of potatoes; cover potatoes in sufcient water and bring to boil; when the water boils, reduce heat and keep simmering for 20 minutes; drain potatoes and serve. By following this recipe, you get boiled potatoes. You don’t expect fries, or khoresht-e gheymeh bademjan (eggplant stew). The same principle applies to imaging and its use by AI tools. Yet, this is often not the case. Imaging acquired using one scanner can be quite diferent from one acquired from the same patient using a scanner from another vendor. Worse, quantitative image analysis tools and workfows that purportedly perform the same analysis may produce markedly diferent results for the same image [1]. Such difculties prevent clinical translation of potentially useful AI tools and quantitative imaging biomarkers [2, 3].

Details

Original languageEnglish
Pages (from-to)8842-8843
Number of pages2
JournalEuropean radiology
Volume33
Issue number12
Publication statusE-pub ahead of print - 19 Jul 2023
Peer-reviewedYes
Externally publishedYes

External IDs

Scopus 85165196966

Keywords

Keywords

  • Humans, Reference Standards, Diagnostic Imaging

Library keywords