Unsupervised Lesion Detection in Brain CT using Bayesian Convolutional Autoencoders

Normally, lesions are detected using supervised learning techniques that require labelled training data. We explore the use of Bayesian autoencoders to learn the variability of healthy tissue and detect lesions as unlikely events under the normative model. As a proof-of-concept, we test our method on registered 2D midaxial slices from CT imaging data. Our results indicate that our method achieves best performance in detecting lesions caused by bleeding compared to baselines.

[1]  Brendt Wohlberg,et al.  Detecting anomalous structures by convolutional sparse models , 2015, 2015 International Joint Conference on Neural Networks (IJCNN).

[2]  Ben Glocker,et al.  Implicit Weight Uncertainty in Neural Networks. , 2017 .

[3]  Ben Glocker,et al.  DLTK: State of the Art Reference Implementations for Deep Learning on Medical Images , 2017, ArXiv.

[4]  Koenraad Van Leemput,et al.  Automated segmentation of multiple sclerosis lesions by model outlier detection , 2001, IEEE Transactions on Medical Imaging.

[5]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[6]  Georg Langs,et al.  Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery , 2017, IPMI.

[7]  Bram van Ginneken,et al.  A survey on deep learning in medical image analysis , 2017, Medical Image Anal..

[8]  Max Welling,et al.  Multiplicative Normalizing Flows for Variational Bayesian Neural Networks , 2017, ICML.

[9]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[10]  Sungzoon Cho,et al.  Variational Autoencoder based Anomaly Detection using Reconstruction Probability , 2015 .

[11]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[12]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[13]  Zoubin Ghahramani,et al.  Bayesian Convolutional Neural Networks with Bernoulli Approximate Variational Inference , 2015, ArXiv.