We propose a novel semi-supervised image segmentation method that simultaneously optimizes a supervised segmentation and an unsupervised reconstruction objectives. The reconstruction objective uses an attention mechanism that separates the reconstruction of image areas corresponding to different classes. The proposed approach was evaluated on two applications: brain tumor and white matter hyperintensities segmentation. Our method, trained on unlabeled and a small number of labeled images, outperformed supervised CNNs trained with the same number of images and CNNs pre-trained on unlabeled data. In ablation experiments, we observed that the proposed attention mechanism substantially improves segmentation performance. We explore two multi-task training strategies: joint training and alternating training. Alternating training requires fewer hyperparameters and achieves a better, more stable performance than joint training. Finally, we analyze the features learned by different methods and find that the attention mechanism helps to learn more discriminative features in the deeper layers of encoders.

The choice of features greatly influences the performance of a tissue classification system. Despite this, many systems are built with standard, predefined filter banks that are not optimized for that particular application. Representation learning methods such as restricted Boltzmann machines may outperform these standard filter banks because they learn a feature description directly from the training data. Like many other representation learning methods, restricted Boltzmann machines are unsupervised and are trained with a generative learning objective; this allows them to learn representations from unlabeled data, but does not necessarily produce features that are optimal for classification. In this paper we propose the convolutional classification restricted Boltzmann machine, which combines a generative and a discriminative learning objective. This allows it to learn filters that are good both for describing the training data and for classification. We present experiments with feature learning for lung texture classification and airway detection in CT images. In both applications, a combination of learning objectives outperformed purely discriminative or generative learning, increasing, for instance, the lung tissue classification accuracy by 1 to 8 percentage points. This shows that discriminative learning can help an otherwise unsupervised feature learner to learn filters that are optimized for classification.

Abstract. The classification and registration of incomplete multi-modal medical images, such as multi-sequence MRI with missing sequences, can sometimes be improved by replacing the missing modalities with synthetic data. This may seem counter-intuitive: synthetic data is derived from data that is already available, so it does not add new information. Why can it still improve performance? In this paper we discuss possible explanations. If the synthesis model is more flexible than the classifier, the synthesis model can provide features that the classifier could not have extracted from the original data. In addition, using synthetic information to complete incomplete samples increases the size of the training set.We present experiments with two classifiers, linear support vector machines (SVMs) and random forests, together with two synthesis methods that can replace missing data in an image classification problem: neural networks and restricted Boltzmann machines (RBMs). We used data from the BRATS 2013 brain tumor segmentation challenge, which includes multi-modal MRI scans with T1, T1 post-contrast, T2 and FLAIR sequences. The linear SVMs appear to benefit from the complex transformations offered by the synthesis models, whereas the random forests mostly benefit from having more training data. Training on the hidden representation from the RBM brought the accuracy of the linear SVMs close to that of random forests.

No abstract

Machine learning algorithms can have difficulties adapting to data from different sources, for example from different imaging modalities. We present and analyze three techniques for unsupervised cross-modality feature learning, using a shared autoencoder-like convolutional network that learns a common representation from multi-modal data. We investigate a form of feature normalization, a learning objective that minimizes crossmodality differences, and modality dropout, in which the network is trained with varying subsets of modalities. We measure the same-modality and cross-modality classification accuracies and explore whether the models learn modality-specific or shared features. This paper presents experiments on two public datasets, with knee images from two MRI modalities, provided by the Osteoarthritis Initiative, and brain tumor segmentation on four MRI modalities from the BRATS challenge. All three approaches improved the cross-modality classification accuracy, with modality dropout and per-feature normalization giving the largest improvement. We observed that the networks tend to learn a combination of cross-modality and modality-specific features. Overall, a combination of all three methods produced the most cross-modality features and the highest cross-modality classification accuracy, while maintaining most of the samemodality accuracy.

Abstract. We developed a learning-based question classifier for question answering systems. A question classifier tries to predict the entity type of the possible answers to a given question written in natural language. We extracted several lexical, syntactic and semantic features and examined their usefulness for question classification. Furthermore we developed a weighting approach to combine features based on their importance. Our result on the well-known TREC questions dataset is competitive with the state-of-the-art on this task.

Background In Pompe disease, an inherited metabolic muscle disorder, severe diaphragmatic weakness often occurs. Enzyme replacement treatment is relatively ineffective for respiratory function, possibly because of irreversible damage to the diaphragm early in the disease course. Mildly impaired diaphragmatic function may not be recognized by spirometry, which is commonly used to study respiratory function. In this cross-sectional study, we aimed to identify early signs of diaphragmatic weakness in Pompe patients using chest MRI. Methods Pompe patients covering the spectrum of disease severity, and sex and age matched healthy controls were prospectively included and studied using spirometry-controlled sagittal MR images of both mid-hemidiaphragms during forced inspiration. The motions of the diaphragm and thoracic wall were evaluated by measuring thoracic cranial-caudal and anterior–posterior distance ratios between inspiration and expiration. The diaphragm shape was evaluated by measuring the height of the diaphragm curvature. We used multiple linear regression analysis to compare different groups. Results We included 22 Pompe patients with decreased spirometry results (forced vital capacity in supine position < 80% predicted); 13 Pompe patients with normal spirometry results (forced vital capacity in supine position ≥ 80% predicted) and 18 healthy controls. The mean cranial-caudal ratio was only 1.32 in patients with decreased spirometry results, 1.60 in patients with normal spirometry results and 1.72 in healthy controls (p < 0.001). Anterior–posterior ratios showed no significant differences. The mean height ratios of the diaphragm curvature were 1.41 in patients with decreased spirometry results, 1.08 in patients with normal spirometry results and 0.82 in healthy controls (p = 0.001), indicating an increased curvature of the diaphragm during inspiration in Pompe patients. Conclusions Even in early-stage Pompe disease, when spirometry results are still within normal range, the motion of the diaphragm is already reduced and the shape is more curved during inspiration. MRI can be used to detect early signs of diaphragmatic weakness in patients with Pompe disease, which might help to select patients for early intervention to prevent possible irreversible damage to the diaphragm.

No abstract