Comparison of machine learning and deep learning for view identification from cardiac magnetic resonance images.
Comparison of machine learning and deep learning for view identification from cardiac magnetic resonance images.
复制标题
DOI：
10.1016/j.clinimag.2021.11.013
复制
发表时间：
2022-03
期刊：
影响因子：
2.1
通讯作者：
Patel AR
中科院分区：
文献类型：
--
作者：
Chauhan D;Anyanwu E;Goes J;Besser SA;Anand S;Madduri R;Getty N;Kelle S;Kawaji K;Mor-Avi V;Patel AR
基于人工智能的心脏磁共振 (CMR) 研究解读依赖于准确识别成像平面的能力，这可以通过深度学习 (DL) 和机器学习 (ML) 技术来实现。我们的目的是比较 ML 和 DL 在 CMR 视图分类方面的准确性，并识别算法训练和测试期间的潜在缺陷。深度学习和机器学习算法都能准确地对 CMR 图像进行分类，但在对复杂心脏解剖结构的图像进行分类时，深度学习优于机器学习（95% 和 90%）。要达到这种准确性水平，需要对精心策划的研究队列进行培训。人工智能越来越多地用于帮助解释心脏磁共振 (CMR) 研究。第一步是识别所描绘的成像平面，这可以通过深度学习 (DL) 和经典机器学习 (ML) 技术来实现，无需用户输入。我们的目的是比较 ML 和 DL 在 CMR 视图分类方面的准确性，并识别算法训练和测试期间的潜在缺陷。为了训练我们的深度学习和机器学习算法，我们首先通过回顾性选择 200 个 CMR 案例来建立数据集。这些模型使用两个不同的队列（被动和主动策划）进行训练，并应用数据增强来增强训练。训练完成后，模型将在外部数据集上进行验证，该数据集包含在另一个中心获取的 20 个病例。然后，我们比较了准确性指标和应用类激活映射 (CAM)，以可视化 DL 模型性能。使用被动管理的 CMR 队列训练的 DL 和 ML 模型在验证集上的准确率分别为 99.1% 和 99.3%。然而，当在具有复杂解剖结构的 CMR 病例上进行测试时，两种模型的表现都很差。在对所有 200 个案例（活跃队列）再次训练和测试我们的模型后，对外部数据集的验证分别获得了 95% 和 90% 的准确率。 CAM 分析描绘了热图，证明了仔细整理用于训练的数据集的重要性。深度学习和机器学习模型都可以准确地对 CMR 图像进行分类，但在对具有复杂心脏解剖结构的图像进行分类时，深度学习优于机器学习。在对具有复杂心脏解剖结构的图像进行分类时，深度学习的表现优于机器学习（95% vs 90%）。要达到这种准确性水平，需要对精心策划的研究队列进行培训。
Artificial intelligence based interpretation of cardiac magnetic resonance (CMR) studies relies on the ability to accurately identify the imaging plane, which can be achieved by both deep learning (DL) and machine learning (ML) techniques. We aimed to compare the accuracy of ML and DL for CMR view classification and to identify potential pitfalls during training and testing of the algorithms. Both DL and ML algorithms accurately classified CMR images, but DL outperformed ML (95% and 90%) when classifying images with complex heart anatomy. Reaching this level of accuracy required training on a carefully curated cohort of studies. Artificial intelligence is increasingly utilized to aid in the interpretation of cardiac magnetic resonance (CMR) studies. One of the first steps is the identification of the imaging plane depicted, which can be achieved by both deep learning (DL) and classical machine learning (ML) techniques without user input. We aimed to compare the accuracy of ML and DL for CMR view classification and to identify potential pitfalls during training and testing of the algorithms. To train our DL and ML algorithms, we first established datasets by retrospectively selecting 200 CMR cases. The models were trained using two different cohorts (passively and actively curated) and applied data augmentation to enhance training. Once trained, the models were validated on an external dataset, consisting of 20 cases acquired at another center. We then compared accuracy metrics and applied class activation mapping (CAM) to visualize DL model performance. The DL and ML models trained with the passively-curated CMR cohort were 99.1% and 99.3% accurate on the validation set, respectively. However, when tested on the CMR cases with complex anatomy, both models performed poorly. After training and testing our models again on all 200 cases (active cohort), validation on the external dataset resulted in 95% and 90% accuracy, respectively. The CAM analysis depicted heat maps that demonstrated the importance of carefully curating the datasets to be used for training. Both DL and ML models can accurately classify CMR images, but DL outperformed ML when classifying images with complex heart anatomy. Deep learning outperformed machine learning (95% vs 90%) when classifying images with complex heart anatomy. Reaching this level of accuracy required training on a carefully curated cohort of studies.
登录
查看更多内容
DOI：
10.1148/ryai.2020190043
发表时间：
2020-05-01
期刊：
RADIOLOGY-ARTIFICIAL INTELLIGENCE
影响因子：
--
作者：
Reyes, Mauricio;Meier, Raphael;Wiest, Roland
通讯作者：
Wiest, Roland
影响因子：
4.5
作者：
Thrall, James H.;Li, Xiang;Brink, James
通讯作者：
Brink, James
影响因子：
15.2
作者：
Madani, Ali;Arnaout, Ramy;Arnaout, Rima
通讯作者：
Arnaout, Rima
影响因子：
8.1
作者：
Johnson, Justin M.;Khoshgoftaar, Taghi M.
通讯作者：
Khoshgoftaar, Taghi M.
影响因子：
3
作者：
McNemar, Quinn
通讯作者：
McNemar, Quinn