Different aspects of cognitive functions are affected in patients with Alzheimer’s disease. To date, little is known about the associations between features from brain-imaging and individual Alzheimer’s disease (AD)-related cognitive functional changes. In addition, how these associations differ among different imaging modalities is unclear. Here, we trained and investigated 3D convolutional neural network (CNN) models that predicted sub-scores of the 13-item Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS–Cog13) based on MRI and FDG–PET brain-imaging data. Analysis of the trained network showed that each key ADAS–Cog13 sub-score was associated with a specific set of brain features within an imaging modality. Furthermore, different association patterns were observed in MRI and FDG–PET modalities. According to MRI, cognitive sub-scores were typically associated with structural changes of subcortical regions, including amygdala, hippocampus, and putamen. Comparatively, according to FDG–PET, cognitive functions were typically associated with metabolic changes of cortical regions, including the cingulated gyrus, occipital cortex, middle front gyrus, precuneus cortex, and the cerebellum. These findings brought insights into complex AD etiology and emphasized the importance of investigating different brain-imaging modalities.

中文翻译：

---

3D 卷积神经网络揭示了阿尔茨海默病分项评分的特定模态脑成像预测因子

阿尔茨海默病患者认知功能的不同方面受到影响。迄今为止，人们对大脑成像特征与个体阿尔茨海默病（AD）相关认知功能变化之间的关联知之甚少。此外，这些关联在不同成像方式之间有何不同尚不清楚。在这里，我们训练和研究了 3D 卷积神经网络 (CNN) 模型，该模型根据 MRI 和 FDG-PET 脑成像数据预测 13 项阿尔茨海默病评估量表 - 认知子量表 (ADAS-Cog13) 的子分数。对训练网络的分析表明，每个关键的 ADAS-Cog13 子分数都与成像模式中的一组特定大脑特征相关。此外，在 MRI 和 FDG-PET 模式中观察到不同的关联模式。根据 MRI，认知评分通常与皮质下区域的结构变化相关，包括杏仁核、海马体和壳核。相比之下，根据FDG-PET，认知功能通常与皮质区域的代谢变化相关，包括扣带回、枕叶皮质、中前回、楔前叶皮质和小脑。这些发现使我们对复杂的 AD 病因学有了深入的了解，并强调了研究不同脑成像模式的重要性。