在CT扫描中基于深度学习的脑膜和心室系统的细分在存在异常的情况下
Annika Gerken1, Sina Walluscheck2, Peter Kohlmann3
1Fraunhofer Institute for Digital Medicine MEVIS, Bremen, Germany.
Frontiers in neuroimaging
|August 9, 2023
概括
将脑出血数据添加到深度学习培训中,可以显著改善CT扫描中脑结构的细分. 这种增强的稳定性有利于对正常和病态脑部扫描的分析,有助于临床诊断.
科学领域:
- 医疗成像医学成像
- 人工智能的人工智能
- 神经科学是一个神经科学.
背景情况:
- 准确细分大脑外围膜和脑脊液空间对于分析大脑CT数据至关重要.
- 对于临床诊断应用而言,对解剖学变异和病理的强度至关重要.
研究的目的:
- 调查结合出血训练数据对深度学习细分算法的稳定性的影响.
- 评估算法在正常,出血和瘤大脑CT数据集上的性能.
主要方法:
- 在正常的大脑解剖学数据上训练了一个2D U-Net深度学习模型.
- 第二个模型使用相同的数据和额外的脑出血受试者进行了训练.
- 这两种模型都在单独的正常,出血和瘤 (GLIS-RT数据集) 测试案例上进行了评估.
主要成果:
- 与仅在正常数据上训练的模型相比,使用出血数据的训练显著改善了所有测试集的细分性能.
- 改进的算法获得了高的子中位数得分:0.98 (大脑膜),0.91 (左心室),0.90 (右心室) 在出血测试数据上.
- 在瘤测试组中,Dice的中位数得分为0.96 (腹膜),0.90 (左心室),0.90 (右心室),证明了更好的概括性.
结论:
- 将病理学数据扩展到训练数据集对于提高脑CT细分算法的稳定性至关重要.
- 改进的算法证明了对细分大脑副体和心室系统的稳定性增加,即使对于看不见的异常.
- 进一步扩展训练数据,包括各种病理,可以提高算法的概括性.
相关概念视频
Anatomy of the Brain: Ventricles
4.3K
There are hollow fluid-filled cavities known as ventricles deep inside the human brain. There are two lateral ventricles, one in each cerebral hemisphere, and each has three different projections — the anterior, inferior, and posterior horns visible from the lateral side. A thin membrane called the septum pellucidum separates the two lateral ventricles. The slender third ventricle in the diencephalon is connected to each lateral ventricle via a channel called the interventricular foramen.
4.3K
Imaging Studies for Cardiovascular System V: CT
42
Cardiac computed tomography (CT) scanning is an advanced cardiac imaging technique that utilizes CT technology, with or without intravenous (IV) contrast, to produce accurate cross-sectional virtual slices of specific areas of the heart, coronary circulation, and major blood vessels such as the aorta, pulmonary veins, and arteries. The computer processes these slices to generate three-dimensional images. Multidetector CT (MDCT) is a rapid form of CT scanning that captures multiple slices...
42


