基于生成对抗网络的切片选择性学习用于阿尔茨海默病分类:外部验证的可行性研究
本可行性研究采用切片选择性学习结合边界平衡生成对抗网络(BEGAN)提取特征,训练支持向量机分类器区分阿尔茨海默病(AD)与正常认知(NC)。模型在ADNI数据集训练,在Severance医院数据集外部验证,覆盖后扣带回皮层(PCC)的双切片范围表现最佳;两个独立数据集上的性能无统计学差异(p > 0.05),双切片与单切片输入的性能存在统计学差异(p < 0.05)。
要点
- 覆盖后扣带回皮层(PCC)的双切片范围在切片选择性学习中表现最佳。
- Severance数据集上准确率94.33%、灵敏度91.78%、特异度97.06%;ADNI数据集上准确率94.82%、灵敏度92.11%、特异度97.45%。
- 两个独立数据集性能无统计学差异(p > 0.05);双切片与单切片输入性能有统计学差异(p < 0.05)。
核心结论
- 当选择适当的切片时,模型学习到了AD与NC的泛化特征,可用于外部验证。
- 本研究显示了该模型在多种图像采集环境下测试时性能一致的可行性。
研究设计与样本
- 可行性研究:使用阿尔茨海默病神经影像学倡议(ADNI)数据库的18F-氟脱氧葡萄糖([18F]FDG)PET/CT数据集训练模型,并使用Severance医院数据集进行测试。
- 采用切片选择性学习以降低计算成本并提取无偏特征,使用边界平衡生成对抗网络(BEGAN)提取AD与正常认知(NC)的特征,再用这些特征训练支持向量机(SVM)分类器区分AD与NC。
- 对每个切片进行评估以研究切片性能。
主要发现
- 覆盖后扣带回皮层(PCC)的双切片范围表现最佳。
- 所提出网络在Severance数据集上的准确率、灵敏度和特异度分别为94.33%、91.78%和97.06%。
- 所提出网络在ADNI数据集上的准确率、灵敏度和特异度分别为94.82%、92.11%和97.45%。
- 两个独立数据集上的性能无统计学差异(p > 0.05)。
- 使用双切片与单切片作为输入的性能之间存在统计学差异(p < 0.05)。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
PURPOSE: The aim of this feasibility study was to use slice selective learning using a Generative Adversarial Network for external validation. We aimed to build a model less sensitive to PET imaging acquisition environment, since differences in environments negatively influence network performance. To investigate the slice performance, each slice evaluation was performed. METHODS: We trained our model using a 18F-fluorodeoxyglucose ([18F]FDG) PET/CT dataset obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database and tested the model with a Severance Hospital dataset. We applied slice selective learning to reduce computational cost and to extract unbiased features. We extracted features of Alzheimer's disease (AD) and normal cognitive (NC) condition using a Boundary Equilibrium Generative Adversarial Network (BEGAN) for stable convergence. Then, we utilized these features to train a support vector machine (SVM) classifier to distinguish AD from NC. RESULTS: The slice range that covered the posterior cingulate cortex (PCC) using double slices showed the best performance. The accuracy, sensitivity, and specificity of our proposed network was 94.33%, 91.78%, and 97.06% using the Severance dataset and 94.82%, 92.11%, and 97.45% using the ADNI dataset. The performance on the two independent datasets showed no statistical difference (p > 0.05). Moreover, there was a statistical difference in the performance between using two slices and one slice as input (p < 0.05). CONCLUSIONS: Our model learned the generalized features of AD and NC for external validation when appropriate slices were selected. This study showed the feasibility of this model with consistent performance when tested using datasets acquired from a variety of image-acquisition environments.
引用
- Kim HW, Lee HE, Lee S, et al. Slice-selective learning for Alzheimer's disease classification using a generative adversarial network: a feasibility study of external validation. Eur J Nucl Med Mol Imaging. 2020. PMID: 31980910. doi:10.1007/s00259-019-04676-y
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。