Division of Neuroscience and Behavior, National Institute on Drug Abuse, National Institutes of Health, 11601 Landsdown St., Three White Flint North (3WFN), MSC 6018, Rockville, MD 20852, United States.
Department of Psychiatry, Yale University School of Medicine, 40 Temple Street, New Haven, CT, 06511, United States; Interdepartmental Neuroscience Program, Yale University, New Haven, CT, 06511, United States; Department of Psychology, Yale University, New Haven, CT, 06511, United States.
Psychiatry Res. 2024 Sep;339:115955. doi: 10.1016/j.psychres.2024.115955. Epub 2024 May 20.
The explosion of generative AI offers promise for neuroimaging biomarker development in psychiatry, but effective adoption of AI methods requires clarity with respect to specific applications and challenges. These center on dataset sizes required to robustly train AI models along with feature selection that capture neural signals relevant to symptom and treatment targets. Here we discuss areas where generative AI could improve quantification of robust and reproducible brain-to-symptom associations to inform precision psychiatry applications, especially in the context of drug discovery. Finally, this communication discusses some challenges that need solutions for generative AI models to advance neuroimaging biomarkers in psychiatry.
生成式人工智能的爆炸式发展为精神病学中的神经影像学生物标志物的开发带来了希望，但要有效地采用 AI 方法，就需要明确具体的应用和挑战。这些挑战集中在需要多大的数据集来稳健地训练 AI 模型，以及选择能够捕捉与症状和治疗靶点相关的神经信号的特征。在这里，我们讨论了生成式人工智能可以提高对稳健和可重复的大脑与症状关联的定量分析的领域，以支持精准精神病学的应用，特别是在药物发现的背景下。最后，本通讯讨论了一些生成式人工智能模型在推进精神病学神经影像学生物标志物方面需要解决的挑战。
