Conceitos essenciais
Harmonizing data set terminology between medical and artificial intelligence research fields is crucial for effective communication and collaboration in the rapidly evolving field of medical AI.
Resumo
This narrative review examines the historical evolution of data set terminology in both the medical and artificial intelligence (AI) research fields, highlighting the importance of clear and standardized terminology for effective communication and collaboration in the rapidly growing field of medical AI.
The review begins by tracing the divergent development of data set terminology in the AI and medical research domains. In AI research, the common framework of "training set," "validation set," and "test set" has been widely adopted, while in traditional medical research, the term "validation" has been used to refer to the final testing of a developed model.
The review then explores examples of prominent medical AI studies that have exhibited terminological inconsistencies, leading to potential misunderstandings and methodological discrepancies. To address this issue, the review recommends the adoption of standardized AI-centric terminology, such as "training set," "validation (or tuning) set," and "test set," along with explicit definitions of these terms in each research publication.
Additionally, the review delves into the various categories of test sets used in AI evaluation, including internal testing (random splitting, cross-validation, and leave-one-out) and external testing (temporal and geographic sets). Understanding these test set classifications is crucial for assessing the robustness and generalizability of AI applications in medicine.
The harmonization of data set terminology between medical and AI research is critical for advancing the field of medical AI. By adopting standardized terminologies and ensuring their clear definition in research publications, the review aims to foster more effective communication and collaboration across disciplines, ultimately contributing to the ethical and effective deployment of AI in clinical settings.
Estatísticas
"Medicine and artificial intelligence (AI) engineering represent two distinct fields each with decades of published history."
"The rapid convergence of AI and medicine has led to significant advancements, yet it has also introduced ambiguity regarding data set terms common to both fields, potentially leading to miscommunication and methodological discrepancies."
"This review traces the divergent evolution of terms for data sets and their impact."
"This review clarifies existing literature to provide a comprehensive understanding of these classifications and their implications in AI evaluation."
Citações
"Foremost among these challenges is the confusion arising from the overlapping and often contradictory data set terminologies used in the medical and AI research fields, consequently impacting the fledgling field of medical AI."
"Such terminological overlaps extend beyond academic concerns; they have practical implications for the design, interpretation, and communication of research in medical AI."
"The harmonization of data set terminology between medical and AI research is critical for advancing the field of medical AI."