Ensembles of Multiple Models and Architectures for Robust Brain Tumour Segmentation

By Konstantinos Kamnitsas, Wenjia Bai, Enzo Ferrante, Steven G. McDonagh, Matthew Sinclair, Nick Pawlowski, Martin Rajchl, Matthew C. H. Lee, Bernhard Kainz, Daniel Rueckert and 1 more. BrainLes@MICCAI, 2017

460 citations recorded by the source corpus.

Deep learning approaches such as convolutional neural nets have consistently outperformed previous methods on challenging tasks such as dense, semantic segmentation. However, the various proposed networks perform differently, with behaviour largely influenced by architectural choices and training settings. This paper explores Ensembles of Multiple Models and Architectures (EMMA) for robust performance through aggregation of predictions from a wide range of methods. The approach reduces the influence of the meta-parameters of individual models and the risk of overfitting the configuration to a particular database. EMMA can be seen as an unbiased, generic deep learning model which is shown to yield excellent performance, winning the first position in the BRATS 2017 competition among 50+ participating teams.

Subfields: Computer Vision and Pattern Recognition, Neurology, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/9b6f5e4cb8f2163fe400f1315bca976623e33b28

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Neurology

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

No New-Net (2018, methodology): “Kamnitsas et al. [9], who were the clear winner of the challenge, created an ensemble by combining three diﬀerent network architectures, namely 3D FCN [12], 3D U-Net [15,14] and DeepMedic [5], trained with diﬀerent loss functions (dice loss [16,17] and crossentropy) and diﬀerent normalization…”
REFUGE Challenge: A Unified Framework for Evaluating Automated Methods for Glaucoma Assessment from Fundus Photographs (2019, unknown): “Combining the outcomes of multiple models is a common practice in challenges as it allows to achieve (sometimes marginal) quantitative improvements that can eventually ensure higher positions in the final rankings (Kaggle, 2015; Kamnitsas et al., 2017).”
Attention U-Net: Learning Where to Look for the Pancreas (2018, methodology): “In particular, fully convolutional networks (FCN) [17] such as U-Net [22], DeepMedic [13] and holistically nested networks [15, 31] have been shown to achieve robust and accurate performance in various tasks including cardiac MR [3], brain tumours [12] and abdominal CT [24, 25] image segmentation ta”
Deep Learning for Cardiac Image Segmentation: A Review (2019, methodology): “• Dropout (82): Dropout is a regularization method that randomly drops some units from the neural network during training, encouraging the network to learn a sparse representation; • Ensemble learning: Ensemble learning is a type of machine learning algorithms that combine multiple trained models to”
Deep learning based brain tumor segmentation: a survey (2020, background): “CNN Single-Path [26], [27], [28], [29], [30], [31], [32] Multi-Path [33], [34], [35], [36], [37], [38], [39], [40], [41], [42] FCN Vanilla FCN [43], [44], [28], [45], [46], [47], [48], [49], [50] U/V-Net Based [51], [52], [53], [54], [55], [56], [57], [58], [59], [60] Cascaded CNN [61], [62], [63], ”
Fully convolutional multi‐scale residual DenseNets for cardiac segmentation and automated cardiac diagnosis using ensemble of classifiers (2018, background): “…(2015)) have achieved remarkable success in segmentation of various structures in heart (Dou et al. (2016)), brain lesions (Havaei et al. (2017); Kamnitsas et al. (2017); Pereira et al. (2016)), liver lesions (Christ et al. (2016); Dou et al. (2016); Ben-Cohen et al. (2015); Deng and Du (2008))…”
Attention Gated Networks: Learning to Leverage Salient Regions in Medical Images (2018, methodology): “CNNs are now the state-of-the-art method for many tasks including classiﬁ-cation , localisation and segmentation (Bai et al., 2017; Kamnitsas et al., 2017, 2018; Lee et al., 2015; Litjens et al., 2017; Long et al., 2015; Ronneberger et al., 2015; Roth et al., 2017, 2018; Xie and Tu, 2015; Zaharchuk…”
Confidence Calibration and Predictive Uncertainty Estimation for Deep Medical Image Segmentation (2019, background): “to improve classification performance of deep NNs in several applications including medical image segmentation [24], [25].”
3D MRI brain tumor segmentation using autoencoder regularization (2018, methodology): “to regularize the shared encoder. At inference time, only the main segmentation encode-decoder part is used. 2 Related work Last year, BraTS 2017, top performing submissions included Kamnitsas et al. [13] who proposed to ensemble several models for robust segmentation (EMMA), and Wang et al. [21] wh”
Identifying the Best Machine Learning Algorithms for Brain Tumor Segmentation, Progression Assessment, and Overall Survival Prediction in the BRATS Challenge (2018, background): “S eg m en ta tio n 1 biomedia1 Konstantinos Kamnitsas Imperial College London, UK [33] 2 UCL-TIG Guotai Wang University College London (UCL), UK [55] 3 (tie) MIC DKFZ Fabian Isensee Division of Medical Image Computing, German Cancer Research Center (DKFZ), Heidelberg, Germany [29]”
A review: Deep learning for medical image segmentation using multi-modality fusion (2019, methodology): “[54] [13] [36] [77] [35] used the N4ITK method to correct the distortion of MRI data.”