Department of Psychiatry, University of Geneva, Geneva, Switzerland.
Medical Direction, University of Geneva Hospitals, Geneva, Switzerland.
Eur Radiol. 2022 Nov;32(11):7833-7842. doi: 10.1007/s00330-022-08798-0. Epub 2022 Apr 29.
Established visual brain MRI markers for dementia include hippocampal atrophy (mesio-temporal atrophy MTA), white matter lesions (Fazekas score), and number of cerebral microbleeds (CMBs). We assessed whether novel quantitative, artificial intelligence (AI)-based volumetric scores provide additional value in predicting subsequent cognitive decline in elderly controls.
A prospective study including 80 individuals (46 females, mean age 73.4 ± 3.5 years). 3T MR imaging was performed at baseline. Extensive neuropsychological assessment was performed at baseline and at 4.5-year follow-up. AI-based volumetric scores were derived from 3DT1: Alzheimer Disease Resemblance Atrophy Index (AD-RAI), Brain Age Gap Estimate (BrainAGE), and normal pressure hydrocephalus (NPH) index. Analyses included regression models between cognitive scores and imaging markers.
AD-RAI score at baseline was associated with Corsi (visuospatial memory) decline (10.6% of cognitive variability in multiple regression models). After inclusion of MTA, CMB, and Fazekas scores simultaneously, the AD-RAI score remained as the sole valid predictor of the cognitive outcome explaining 16.7% of its variability. Its percentage reached 21.4% when amyloid positivity was considered an additional explanatory factor. BrainAGE score was associated with Trail Making B (executive functions) decrease (8.5% of cognitive variability). Among the conventional MRI markers, only the Fazekas score at baseline was positively related to the cognitive outcome (8.7% of cognitive variability). The addition of the BrainAGE score as an independent variable significantly increased the percentage of cognitive variability explained by the regression model (from 8.7 to 14%). The addition of amyloid positivity led to a further increase in this percentage reaching 21.8%.
The AI-based AD-RAI index and BrainAGE scores have limited but significant added value in predicting the subsequent cognitive decline in elderly controls when compared to the established visual MRI markers of brain aging, notably MTA, Fazekas score, and number of CMBs.
• AD-RAI score at baseline was associated with Corsi score (visuospatial memory) decline. • BrainAGE score was associated with Trail Making B (executive functions) decrease. • AD-RAI index and BrainAGE scores have limited but significant added value in predicting the subsequent cognitive decline in elderly controls when compared to the established visual MRI markers of brain aging, notably MTA, Fazekas score, and number of CMBs.
已确立的用于痴呆的视觉脑 MRI 标志物包括海马萎缩（内侧颞叶萎缩 MTA）、白质病变（Fazekas 评分）和脑微出血数量（CMB）。我们评估了新型定量、基于人工智能（AI）的容积评分是否在预测老年对照组随后的认知下降方面提供了额外价值。
一项前瞻性研究包括 80 名个体（46 名女性，平均年龄 73.4±3.5 岁）。在基线时进行 3T MRI 成像。在基线和 4.5 年随访时进行广泛的神经心理学评估。从 3DT1 中得出基于 AI 的容积评分：阿尔茨海默病相似性萎缩指数（AD-RAI）、脑龄差距估计（BrainAGE）和正常压力脑积水（NPH）指数。分析包括认知评分与影像学标志物之间的回归模型。
基线时的 AD-RAI 评分与 Corsi（视觉空间记忆）下降相关（多元回归模型中认知变异性的 10.6%）。同时纳入 MTA、CMB 和 Fazekas 评分后，AD-RAI 评分仍然是认知结果的唯一有效预测因子，解释了其变异性的 16.7%。当考虑淀粉样蛋白阳性为额外解释因素时，其百分比达到 21.4%。BrainAGE 评分与 Trail Making B（执行功能）下降相关（认知变异性的 8.5%）。在常规 MRI 标志物中，只有基线时的 Fazekas 评分与认知结果呈正相关（认知变异性的 8.7%）。将 BrainAGE 评分作为独立变量加入后，回归模型解释的认知变异性百分比显著增加（从 8.7%增加到 14%）。淀粉样蛋白阳性的加入使这一百分比进一步增加到 21.8%。
与已确立的脑老化视觉 MRI 标志物（尤其是 MTA、Fazekas 评分和 CMB 数量）相比，基于 AI 的 AD-RAI 指数和 BrainAGE 评分在预测老年对照组随后的认知下降方面具有有限但有意义的附加价值。
基线时的 AD-RAI 评分与 Corsi 评分（视觉空间记忆）下降相关。
BrainAGE 评分与 Trail Making B（执行功能）下降相关。
与已确立的脑老化视觉 MRI 标志物（尤其是 MTA、Fazekas 评分和 CMB 数量）相比，基于 AI 的 AD-RAI 指数和 BrainAGE 评分在预测老年对照组随后的认知下降方面具有有限但有意义的附加价值。
