University of Wisconsin-Madison; 1111 Highland Ave, Madison, WI 53705, United States of America.
University Medical Centre Ljubljana; Zaloška cesta 2, 1000 Ljubljana, Slovenia.
Phys Med Biol. 2022 Sep 30;67(19). doi: 10.1088/1361-6560/ac8f10.
. Neuroimaging uncovers important information about disease in the brain. Yet in Alzheimer's disease (AD), there remains a clear clinical need for reliable tools to extract diagnoses from neuroimages. Significant work has been done to develop deep learning (DL) networks using neuroimaging for AD diagnosis. However, no particular model has emerged as optimal. Due to a lack of direct comparisons and evaluations on independent data, there is no consensus on which modality is best for diagnostic models or whether longitudinal information enhances performance. The purpose of this work was (1) to develop a generalizable DL model to distinguish neuroimaging scans of AD patients from controls and (2) to evaluate the influence of imaging modality and longitudinal data on performance.. We trained a 2-class convolutional neural network (CNN) with and without a cascaded recurrent neural network (RNN). We used datasets of 772 ( = 364,= 408) 3DF-FDG PET scans and 780 ( = 280,= 500) T1-weighted volumetric-3D MR images (containing 131 and 144 patients with multiple timepoints) from the Alzheimer's Disease Neuroimaging Initiative, plus an independent set of 104 ( = 63, = 41)F-FDG PET scans (one per patient) for validation.. ROC analysis showed that PET-trained models outperformed MRI-trained, achieving maximum AUC with the CNN + RNN model of 0.93 ± 0.08, with accuracy 82.5 ± 8.9%. Adding longitudinal information offered significant improvement to performance onF-FDG PET, but not on T1-MRI. CNN model validation with an independentF-FDG PET dataset achieved AUC of 0.99. Layer-wise relevance propagation heatmaps added CNN interpretability.. The development of a high-performing tool for AD diagnosis, with the direct evaluation of key influences, reveals the advantage of usingF-FDG PET and longitudinal data over MRI and single timepoint analysis. This has significant implications for the potential of neuroimaging for future research on AD diagnosis and clinical management of suspected AD patients.
神经影像学揭示了大脑疾病的重要信息。然而，在阿尔茨海默病（AD）中，仍然需要可靠的工具从神经图像中提取诊断。已经做了大量工作来开发使用神经影像学进行 AD 诊断的深度学习（DL）网络。然而，没有一个特定的模型脱颖而出成为最佳模型。由于缺乏对独立数据的直接比较和评估，因此对于哪种模态最适合诊断模型，或者纵向信息是否提高性能，尚无共识。这项工作的目的是：(1) 开发一种可推广的 DL 模型来区分 AD 患者和对照组的神经影像学扫描；(2) 评估成像模态和纵向数据对性能的影响。我们使用了来自阿尔茨海默病神经影像学倡议的 772 名（= 364 名，= 408 名）3DF-FDG PET 扫描和 780 名（= 280 名，= 500 名）T1 加权容积 3D-MR 图像（包含 131 名和 144 名多名时间点的患者）数据集，以及阿尔茨海默病神经影像学倡议的一个独立的 104 名（= 63 名，= 41 名）F-FDG PET 扫描数据集（每位患者一个）用于验证。ROC 分析表明，经过 PET 训练的模型优于经过 MRI 训练的模型，使用 CNN+RNN 模型的最大 AUC 为 0.93±0.08，准确率为 82.5±8.9%。添加纵向信息可显著提高 F-FDG PET 的性能，但对 T1-MRI 则不然。使用独立的 F-FDG PET 数据集对 CNN 进行验证，AUC 为 0.99。层相关传播热图增加了 CNN 的可解释性。开发一种用于 AD 诊断的高性能工具，并直接评估关键影响因素，这表明与 MRI 和单点分析相比，使用 F-FDG PET 和纵向数据具有优势。这对神经影像学在未来 AD 诊断研究和疑似 AD 患者的临床管理中的应用具有重要意义。
