Faculty, Staff and Student Publications
Publication Date
7-1-2024
Journal
Advances in Radiation Oncology
Abstract
Historically, clinician-derived contouring of tumors and healthy tissues has been crucial for radiation therapy (RT) planning. In recent years, advances in artificial intelligence (AI), predominantly in deep learning (DL), have rapidly improved automated contouring for RT applications, particularly for routine organs-at-risk.1, 2, 3 Despite research efforts actively promoting its broader acceptance, clinical adoption of auto-contouring is not yet standard practice.
Notably, within several AI communities, there has been growing enthusiasm to shift from conventional “model-centric” AI approaches (ie, improving a model while keeping the data fixed), to “data-centric” AI approaches (ie, improving the data while keeping a model fixed).4 Although balancing both approaches is typically ideal for crafting the optimal solution for specific-use cases, most research in RT auto-contouring has prioritized algorithmic modifications aimed at enhancing quantitative contouring performance based on geometric (ie, structural overlap) indices5—a clear testament to the “model-centric” AI paradigm.
In this editorial, aimed at clinician end-users and multidisciplinary research teams, we harmonize key insights in contemporary RT auto-contouring algorithmic development to promote the adoption of data-centric AI frameworks for impactful future research directions that would further facilitate clinical acceptance. Of note, the discussion herein draws primarily from literature related to head and neck cancer (HNC), showcasing it as a representative example of a complex disease site. However, these insights apply broadly to auto-contouring across disease sites.
Included in
Bioinformatics Commons, Biomedical Informatics Commons, Medical Sciences Commons, Oncology Commons
Comments
PMID: 38799110