📄 中文摘要
在数据有限且难以获取的情况下，深度学习的应用面临挑战，神经影像学在早期帕金森病检测中尤为明显。该研究聚焦于利用静息态功能磁共振成像（fMRI）数据，分析早期帕金森病的检测问题。研究使用了来自40名受试者的fMRI数据，其中包括20例早期帕金森病患者和20名健康对照者。通过对ImageNet预训练的卷积神经网络进行微调，并在两种不同的数据划分策略下进行评估，结果显示常用的图像级评估方法在此类数据稀缺的情况下存在局限性。该研究为在极端数据稀缺条件下的机器学习提供了新的视角。
在极端数据稀缺下学习：基于受试者的轻量级卷积神经网络在 fMRI 早期帕金森病检测中的评估
📄 English Summary
Learning Under Extreme Data Scarcity: Subject-Level Evaluation of Lightweight CNNs for fMRI-Based Prodromal Parkinsons Detection
This study addresses the challenges of applying deep learning in scenarios where data is limited and difficult to obtain, particularly in neuroimaging for prodromal Parkinson's disease. It focuses on detecting prodromal Parkinson's using resting-state fMRI data. The research utilizes fMRI data from 40 subjects, including 20 prodromal Parkinson's cases and 20 healthy controls. ImageNet-pretrained convolutional neural networks are fine-tuned and evaluated under two different data partitioning strategies. Results indicate that commonly used image-level evaluation methods have limitations in the context of extreme data scarcity. This work provides new insights into machine learning under such conditions.