Prince Eric W, Mirsky David M, Hankinson Todd C, Görg Carsten
University of Colorado Anschutz Medical Campus, Aurora, CO.
Colorado School of Public Health, Aurora, CO.
AMIA Annu Symp Proc. 2025 May 22;2024:930-939. eCollection 2024.
This research explores the integration of Artificial Intelligence (AI) into clinical decision-making in pediatric brain tumor care, specifically Adamantinomatous Craniopharyngioma (ACP). We present a user-centered design approach to introducing AI tools into clinical workflows to support decision-making in managing Central Nervous System tumors. We conducted a controlled experiment with six clinical experts to explore the hypothesis that AI integrated into clinical contexts can improve the radiographic interpretation of ACP. We found that AI assistance reduced task difficulty and enhanced clinical efficiency; we also discovered variations in user behavior during the annotation process. We identified multiple challenges, including the interpretive complexity of radiographic images and increased disagreements among clinicians when AI was employed. Our study underscores the importance of a nuanced understanding of clinician experiences for successful AI integration into a high-stakes clinical workflow.
本研究探讨了将人工智能（AI）整合到小儿脑肿瘤护理，特别是成釉细胞瘤型颅咽管瘤（ACP）的临床决策中。我们提出了一种以用户为中心的设计方法，将人工智能工具引入临床工作流程，以支持中枢神经系统肿瘤管理中的决策。我们对六位临床专家进行了一项对照实验，以探讨将人工智能整合到临床环境中可以改善ACP影像学解释的假设。我们发现，人工智能辅助降低了任务难度，提高了临床效率；我们还发现了注释过程中用户行为的差异。我们确定了多个挑战，包括影像学图像的解释复杂性以及使用人工智能时临床医生之间分歧的增加。我们的研究强调了对临床医生经验进行细致入微的理解对于成功将人工智能整合到高风险临床工作流程中的重要性。
