Class imbalance presents a major hurdle in the application of classification methods. A commonly taken approach is to learn ensembles of classifiers using rebalanced data. Examples include bootstrap averaging (bagging) combined with either undersampling or oversampling of the minority class examples. However, rebalancing methods entail asymmetric changes to the examples of different classes, which in turn can introduce their own biases. Furthermore, these methods often require specifying the performance measure of interest a priori, i.e., before learning. An alternative is to employ the threshold moving technique, which applies a threshold to the continuous output of a model, offering the possibility to adapt to a performance measure a posteriori, i.e., a plug-in method. Surprisingly, little attention has been paid to this combination of a bagging ensemble and threshold-moving. In this paper, we study this combination and demonstrate its competitiveness. Contrary to the other resampling methods, we preserve the natural class distribution of the data resulting in well-calibrated posterior probabilities. Additionally, we extend the proposed method to handle multiclass data. We validated our method on binary and multiclass benchmark data sets by using both, decision trees and neural networks as base classifiers. We perform analyses that provide insights into the proposed method.

Most seizure forecasting employs statistical learning techniques that lack a representation of the network interactions that give rise to seizures. We present an epilepsy network emulator (ENE) that uses a network of interconnected phase-locked loops (PLLs) to model synchronous, circuit-level oscillations between electrocorticography (ECoG) electrodes. Using ECoG data from a canine-epilepsy model (Davis et al. 2011) and a physiological entropy measure (approximate entropy or ApEn, Pincus 1995), we demonstrate the entropy of the emulator phases increases dramatically during ictal periods across all ECoG recording sites and across all animals in the sample. Further, this increase precedes the observable voltage spikes that characterize seizure activity in the ECoG data. These results suggest that the ENE is sensitive to phase-domain information in the neural circuits measured by ECoG and that an increase in the entropy of this measure coincides with increasing likelihood of seizure activity. Understanding this unpredictable phase-domain electrical activity present in ECoG recordings may provide a target for seizure detection and feedback control.

Despite the state-of-the-art performance for medical image segmentation, deep convolutional neural networks (CNNs) have rarely provided uncertainty estimations regarding their segmentation outputs, e.g., model (epistemic) and image-based (aleatoric) uncertainties. In this work, we analyze these different types of uncertainties for CNN-based 2D and 3D medical image segmentation tasks at both pixel level and structure level. We additionally propose a test-time augmentation-based aleatoric uncertainty to analyze the effect of different transformations of the input image on the segmentation output. Test-time augmentation has been previously used to improve segmentation accuracy, yet not been formulated in a consistent mathematical framework. Hence, we also propose a theoretical formulation of test-time augmentation, where a distribution of the prediction is estimated by Monte Carlo simulation with prior distributions of parameters in an image acquisition model that involves image transformations and noise. We compare and combine our proposed aleatoric uncertainty with model uncertainty. Experiments with segmentation of fetal brains and brain tumors from 2D and 3D Magnetic Resonance Images (MRI) showed that 1) the test-time augmentation-based aleatoric uncertainty provides a better uncertainty estimation than calculating the test-time dropout-based model uncertainty alone and helps to reduce overconfident incorrect predictions, and 2) our test-time augmentation outperforms a single-prediction baseline and dropout-based multiple predictions.

The first year of life is the most dynamic and perhaps the most critical phase of postnatal brain development. The ability to accurately measure structure changes is critical in early brain development study, which highly relies on the performances of image segmentation and registration techniques. However, either infant image segmentation or registration, if deployed independently, encounters much more challenges than segmentation/registration of adult brains due to dynamic appearance change with rapid brain development. In fact, image segmentation and registration of infant images can assists each other to overcome the above challenges by using the growth trajectories (i.e., temporal correspondences) learned from a large set of training subjects with complete longitudinal data. Specifically, a one-year-old image with ground-truth tissue segmentation can be first set as the reference domain. Then, to register the infant image of a new subject at earlier age, we can estimate its tissue probability maps, i.e., with sparse patch-based multi-atlas label fusion technique, where only the training images at the respective age are considered as atlases since they have similar image appearance. Next, these probability maps can be fused as a good initialization to guide the level set segmentation. Thus, image registration between the new infant image and the reference image is free of difficulty of appearance changes, by establishing correspondences upon the reasonably segmented images. Importantly, the segmentation of new infant image can be further enhanced by propagating the much more reliable label fusion heuristics at the reference domain to the corresponding location of the new infant image via the learned growth trajectories, which brings image segmentation and registration to assist each other. It is worth noting that our joint segmentation and registration framework is also flexible to handle the registration of any two infant images even with significant age gap in the first year of life, by linking their joint segmentation and registration through the reference domain. Thus, our proposed joint segmentation and registration method is scalable to various registration tasks in early brain development studies. Promising segmentation and registration results have been achieved for infant brain MR images aged from 2-week-old to 1-year-old, indicating the applicability of our method in early brain development study.

Detecting COVID-19 in computed tomography (CT) or radiography images has been proposed as a supplement to the RT-PCR test. We compare slice-based (2D) and volume-based (3D) approaches to this problem and propose a deep learning ensemble, called IST-CovNet, combining the best 2D and 3D systems with novel preprocessing and attention modules and the use of a bidirectional Long Short-Term Memory model for combining slice-level decisions. The proposed ensemble obtains 90.80% accuracy and 0.95 AUC score overall on the newly collected IST-C dataset in detecting COVID-19 among normal controls and other types of lung pathologies; and 93.69% accuracy and 0.99 AUC score on the publicly available MosMedData dataset that consists of COVID-19 scans and normal controls only. The system also obtains state-of-art results (90.16% accuracy and 0.94 AUC) on the COVID-CT-MD dataset which is only used for testing. The system is deployed at Istanbul University Cerrahpaşa School of Medicine where it is used to automatically screen CT scans of patients, while waiting for RT-PCR tests or radiologist evaluation.

In1 recent years, there has been a great interest in prostate segmentation, which is a important and challenging task for CT image guided radiotherapy. In this paper, a learning-based segmentation method via joint transductive feature selection and transductive regression is presented, which incorporates the physician's simple manual specification (only taking a few seconds), to aid accurate segmentation, especially for the case with large irregular prostate motion. More specifically, for the current treatment image, experienced physician is first allowed to manually assign the labels for a small subset of prostate and non-prostate voxels, especially in the first and last slices of the prostate regions. Then, the proposed method follows the two step: in prostate-likelihood estimation step, two novel algorithms: tLasso and wLapRLS, will be sequentially employed for transductive feature selection and transductive regression, respectively, aiming to generate the prostate-likelihood map. In multi-atlases based label fusion step, the final segmentation result will be obtained according to the corresponding prostate-likelihood map and the previous images of the same patient. The proposed method has been substantially evaluated on a real prostate CT dataset including 24 patients with 330 CT images, and compared with several state-of-the-art methods. Experimental results show that the proposed method outperforms the state-of-the-arts in terms of higher Dice ratio, higher true positive fraction, and lower centroid distances. Also, the results demonstrate that simple manual specification can help improve the segmentation performance, which is clinically feasible in real practice.

Prediction of individual mobility is crucial in human mobility related applications. Whereas, existing research on individual mobility prediction mainly focuses on next location prediction and short-term dependencies between traveling locations. Long-term location sequence prediction is of great importance for long-time traffic planning and location advertising, and long-term dependencies exist as individual mobility regularity typically occurs daily and weekly. This paper proposes a novel hierarchical temporal attention-based LSTM encoder-decoder model for individual location sequence prediction. The proposed hierarchical attention mechanism captures both long-term and short-term dependencies underlying in individual longitudinal trajectories, and uncovers frequential and periodical mobility patterns in an interpretable manner by incorporating the calendar cycle of individual travel regularities into location prediction. More specifically, the hierarchical attention consists of local temporal attention to identify highly related locations in each day, and global temporal attention to discern important travel regularities over a week. Experiments on individual trajectory datasets with varying degree of traveling uncertainty demonstrate that our method outperforms four baseline methods on three evaluation metrics. In addition, we explore the interpretability of the proposed model in understanding individual daily, and weekly mobility patterns by visualizing the temporal attention weights and frequent traveling patterns associated with locations.