Deep learning

In-depth tissue profiling using multiplexed immunohistochemical consecutive staining on single slide

Despite remarkable recent achievements of immunotherapy strategies in cancer treatment, clinical responses remain limited to subsets of patients. Predictive markers of disease course and response to immunotherapy are urgently needed. Recent results have revealed the potential predictive value of immune cell phenotype and spatial distribution at the tumor site, prompting the need for multidimensional immunohistochemical analyses of tumor tissues. To address this need, we developed a sample-sparing, highly multiplexed immunohistochemistry technique based on iterative cycles of tagging, image scanning, and destaining of chromogenic substrate on a single slide. This assay, in combination with a newly developed automated digital landscaping solution, democratizes access to high-dimensional immunohistochemical analyses by capturing the complexity of the immunome using routine pathology standards. Applications of the method extend beyond cancer to screen and validate comprehensive panels of tissue-based prognostic and predictive markers, perform in-depth in situ monitoring of therapies, and identify targets of disease.

PEDLA: predicting enhancers with a deep learning-based algorithmic framework

Transcriptional enhancers are non-coding segments of DNA that play a central role in the spatiotemporal regulation of gene expression programs. However, systematically and precisely predicting enhancers remain a major challenge. Although existing methods have achieved some success in enhancer prediction, they still suffer from many issues. We developed a deep learning-based algorithmic framework named PEDLA (https://github.com/wenjiegroup/PEDLA), which can directly learn an enhancer predictor from massively heterogeneous data and generalize in ways that are mostly consistent across various cell types/tissues. We first trained PEDLA with 1,114-dimensional heterogeneous features in H1 cells, and demonstrated that PEDLA framework integrates diverse heterogeneous features and gives state-of-the-art performance relative to five existing methods for enhancer prediction. We further extended PEDLA to iteratively learn from 22 training cell types/tissues. Our results showed that PEDLA manifested superior performance consistency in both training and independent test sets. On average, PEDLA achieved 95.0% accuracy and a 96.8% geometric mean (GM) of sensitivity and specificity across 22 training cell types/tissues, as well as 95.7% accuracy and a 96.8% GM across 20 independent test cell types/tissues. Together, our work illustrates the power of harnessing state-of-the-art deep learning techniques to consistently identify regulatory elements at a genome-wide scale from massively heterogeneous data across diverse cell types/tissues.

DanQ: a hybrid convolutional and recurrent deep neural network for quantifying the function of DNA sequences

Modeling the properties and functions of DNA sequences is an important, but challenging task in the broad field of genomics. This task is particularly difficult for non-coding DNA, the vast majority of which is still poorly understood in terms of function. A powerful predictive model for the function of non-coding DNA can have enormous benefit for both basic science and translational research because over 98% of the human genome is non-coding and 93% of disease-associated variants lie in these regions. To address this need, we propose DanQ, a novel hybrid convolutional and bi-directional long short-term memory recurrent neural network framework for predicting non-coding function de novo from sequence. In the DanQ model, the convolution layer captures regulatory motifs, while the recurrent layer captures long-term dependencies between the motifs in order to learn a regulatory 'grammar' to improve predictions. DanQ improves considerably upon other models across several metrics. For some regulatory markers, DanQ can achieve over a 50% relative improvement in the area under the precision-recall curve metric compared to related models. We have made the source code available at the github repository http://github.com/uci-cbcl/DanQ.

Fault Diagnosis for Rotating Machinery Using Vibration Measurement Deep Statistical Feature Learning

Fault diagnosis is important for the maintenance of rotating machinery. The detection of faults and fault patterns is a challenging part of machinery fault diagnosis. To tackle this problem, a model for deep statistical feature learning from vibration measurements of rotating machinery is presented in this paper. Vibration sensor signals collected from rotating mechanical systems are represented in the time, frequency, and time-frequency domains, each of which is then used to produce a statistical feature set. For learning statistical features, real-value Gaussian-Bernoulli restricted Boltzmann machines (GRBMs) are stacked to develop a Gaussian-Bernoulli deep Boltzmann machine (GDBM). The suggested approach is applied as a deep statistical feature learning tool for both gearbox and bearing systems. The fault classification performances in experiments using this approach are 95.17% for the gearbox, and 91.75% for the bearing system. The proposed approach is compared to such standard methods as a support vector machine, GRBM and a combination model. In experiments, the best fault classification rate was detected using the proposed model. The results show that deep learning with statistical feature extraction has an essential improvement potential for diagnosing rotating machinery faults.

Serial grouping of 2D-image regions with object-based attention in humans

After an initial stage of local analysis within the retina and early visual pathways, the human visual system creates a structured representation of the visual scene by co-selecting image elements that are part of behaviorally relevant objects. The mechanisms underlying this perceptual organization process are only partially understood. We here investigate the time-course of perceptual grouping of two-dimensional image-regions by measuring the reaction times of human participants and report that it is associated with the gradual spread of object-based attention. Attention spreads fastest over large and homogeneous areas and is slowed down at locations that require small-scale processing. We find that the time-course of the object-based selection process is well explained by a 'growth-cone' model, which selects surface elements in an incremental, scale-dependent manner. We discuss how the visual cortical hierarchy can implement this scale-dependent spread of object-based attention, leveraging the different receptive field sizes in distinct cortical areas.

DOI:http://dx.doi.org/10.7554/eLife.14320.001

When we look at an object, we perceive it as a whole. However, this is not how the brain processes objects. Instead, cells at early stages of the visual system respond selectively to single features of the object, such as edges. Moreover, each cell responds to its target feature in only a small region of space known as its receptive field. At higher levels of the visual system, cells respond to more complex features: angles rather than edges, for example. The receptive fields of the cells are also larger. For us to see an object, the brain must therefore 'stitch' together diverse features into a unified impression.

This process is termed perceptual grouping. But how does it work? Jeurissen et al. hypothesized that this process depends on the visual system’s attention spreading over a region in the image occupied by an object, and that the speed of the process will depend on the size of the receptive fields involved. If an image region is narrow, the visual system must recruit cells with small receptive fields to process the individual features. Grouping will therefore be slow. By contrast, if the object consists of large uniform areas lacking in detail, grouping should be fast. These assumptions give rise to a model called the “growth-conemodel”, which makes a number of specific predictions about reaction times during perceptual grouping.

Jeurissen et al. tested the growth-cone model’s predictions by measuring the speed of perceptual grouping in 160 human volunteers. These volunteers looked at an image made up of two simple shapes, and reported whether two dots fell on the same or different shapes. The results supported the growth-cone model. People were able to group large and uniform areas quickly, but were slower for narrow areas. Grouping also took more time when the distance between the dots increased. Hence, perceptual grouping of everyday objects calls on a step-by-step process that resembles solving a small maze.

The results also revealed that perceptual grouping of simple shapes relies on the spreading of visual attention over the relevant object. Furthermore, the data support the hypothesis that perceptual grouping makes use of the different sizes of receptive fields at various levels of the visual system. Further research will be needed to translate these findings to the more complex natural scenes we encounter in our daily lives.

DOI:http://dx.doi.org/10.7554/eLife.14320.002

Comparison of deep neural networks to spatio-temporal cortical dynamics of human visual object recognition reveals hierarchical correspondence

The complex multi-stage architecture of cortical visual pathways provides the neural basis for efficient visual object recognition in humans. However, the stage-wise computations therein remain poorly understood. Here, we compared temporal (magnetoencephalography) and spatial (functional MRI) visual brain representations with representations in an artificial deep neural network (DNN) tuned to the statistics of real-world visual recognition. We showed that the DNN captured the stages of human visual processing in both time and space from early visual areas towards the dorsal and ventral streams. Further investigation of crucial DNN parameters revealed that while model architecture was important, training on real-world categorization was necessary to enforce spatio-temporal hierarchical relationships with the brain. Together our results provide an algorithmically informed view on the spatio-temporal dynamics of visual object recognition in the human visual brain.

Genome-wide prediction using Bayesian additive regression trees

BACKGROUND

The goal of genome-wide prediction (GWP) is to predict phenotypes based on marker genotypes, often obtained through single nucleotide polymorphism (SNP) chips. The major problem with GWP is high-dimensional data from many thousands of SNPs scored on several thousands of individuals. A large number of methods have been developed for GWP, which are mostly parametric methods that assume statistical linearity and only additive genetic effects. The Bayesian additive regression trees (BART) method was recently proposed and is based on the sum of nonparametric regression trees with the priors being used to regularize the parameters. Each regression tree is based on a recursive binary partitioning of the predictor space that approximates an unknown function, which will automatically model nonlinearities within SNPs (dominance) and interactions between SNPs (epistasis). In this study, we introduced BART and compared its predictive performance with that of the LASSO, Bayesian LASSO (BLASSO), genomic best linear unbiased prediction (GBLUP), reproducing kernel Hilbert space (RKHS) regression and random forest (RF) methods.

RESULTS

Tests on the QTLMAS2010 simulated data, which are mainly based on additive genetic effects, show that cross-validated optimization of BART provides a smaller prediction error than the RF, BLASSO, GBLUP and RKHS methods, and is almost as accurate as the LASSO method. If dominance and epistasis effects are added to the QTLMAS2010 data, the accuracy of BART relative to the other methods was increased. We also showed that BART can produce importance measures on the SNPs through variable inclusion proportions. In evaluations using real data on pigs, the prediction error was smaller with BART than with the other methods.

CONCLUSIONS

BART was shown to be an accurate method for GWP, in which the regression trees guarantee a very sparse representation of additive and complex non-additive genetic effects. Moreover, the Markov chain Monte Carlo algorithm with Bayesian back-fitting provides a computationally efficient procedure that is suitable for high-dimensional genomic data.

Forecast Modelling via Variations in Binary Image-Encoded Information Exploited by Deep Learning Neural Networks

Traditional forecasting models fit a function approximation from dependent invariables to independent variables. However, they usually get into trouble when date are presented in various formats, such as text, voice and image. This study proposes a novel image-encoded forecasting method that input and output binary digital two-dimensional (2D) images are transformed from decimal data. Omitting any data analysis or cleansing steps for simplicity, all raw variables were selected and converted to binary digital images as the input of a deep learning model, convolutional neural network (CNN). Using shared weights, pooling and multiple-layer back-propagation techniques, the CNN was adopted to locate the nexus among variations in local binary digital images. Due to the computing capability that was originally developed for binary digital bitmap manipulation, this model has significant potential for forecasting with vast volume of data. The model was validated by a power loads predicting dataset from the Global Energy Forecasting Competition 2012.

Extracting a low-dimensional description of multiple gene expression datasets reveals a potential driver for tumor-associated stroma in ovarian cancer

Patterns in expression data conserved across multiple independent disease studies are likely to represent important molecular events underlying the disease. We present the INSPIRE method to infer modules of co-expressed genes and the dependencies among the modules from multiple expression datasets that may contain different sets of genes. We show that INSPIRE infers more accurate models than existing methods to extract low-dimensional representation of expression data. We demonstrate that applying INSPIRE to nine ovarian cancer datasets leads to a new marker and potential driver of tumor-associated stroma, HOPX, followed by experimental validation. The implementation of INSPIRE is available at http://inspire.cs.washington.edu .

Deep Learning Applications for Predicting Pharmacological Properties of Drugs and Drug Repurposing Using Transcriptomic Data

Deep learning is rapidly advancing many areas of science and technology with multiple success stories in image, text, voice and video recognition, robotics, and autonomous driving. In this paper we demonstrate how deep neural networks (DNN) trained on large transcriptional response data sets can classify various drugs to therapeutic categories solely based on their transcriptional profiles. We used the perturbation samples of 678 drugs across A549, MCF-7, and PC-3 cell lines from the LINCS Project and linked those to 12 therapeutic use categories derived from MeSH. To train the DNN, we utilized both gene level transcriptomic data and transcriptomic data processed using a pathway activation scoring algorithm, for a pooled data set of samples perturbed with different concentrations of the drug for 6 and 24 hours. In both pathway and gene level classification, DNN achieved high classification accuracy and convincingly outperformed the support vector machine (SVM) model on every multiclass classification problem, however, models based on pathway level data performed significantly better. For the first time we demonstrate a deep learning neural net trained on transcriptomic data to recognize pharmacological properties of multiple drugs across different biological systems and conditions. We also propose using deep neural net confusion matrices for drug repositioning. This work is a proof of principle for applying deep learning to drug discovery and development.

Model Constrained by Visual Hierarchy Improves Prediction of Neural Responses to Natural Scenes

Accurate estimation of neuronal receptive fields is essential for understanding sensory processing in the early visual system. Yet a full characterization of receptive fields is still incomplete, especially with regard to natural visual stimuli and in complete populations of cortical neurons. While previous work has incorporated known structural properties of the early visual system, such as lateral connectivity, or imposing simple-cell-like receptive field structure, no study has exploited the fact that nearby V1 neurons share common feed-forward input from thalamus and other upstream cortical neurons. We introduce a new method for estimating receptive fields simultaneously for a population of V1 neurons, using a model-based analysis incorporating knowledge of the feed-forward visual hierarchy. We assume that a population of V1 neurons shares a common pool of thalamic inputs, and consists of two layers of simple and complex-like V1 neurons. When fit to recordings of a local population of mouse layer 2/3 V1 neurons, our model offers an accurate description of their response to natural images and significant improvement of prediction power over the current state-of-the-art methods. We show that the responses of a large local population of V1 neurons with locally diverse receptive fields can be described with surprisingly limited number of thalamic inputs, consistent with recent experimental findings. Our structural model not only offers an improved functional characterization of V1 neurons, but also provides a framework for studying the relationship between connectivity and function in visual cortical areas.

Big Data and machine learning in radiation oncology: State of the art and future prospects

Precision medicine relies on an increasing amount of heterogeneous data. Advances in radiation oncology, through the use of CT Scan, dosimetry and imaging performed before each fraction, have generated a considerable flow of data that needs to be integrated. In the same time, Electronic Health Records now provide phenotypic profiles of large cohorts of patients that could be correlated to this information. In this review, we describe methods that could be used to create integrative predictive models in radiation oncology. Potential uses of machine learning methods such as support vector machine, artificial neural networks, and deep learning are also discussed.

Deep learning as a tool for increased accuracy and efficiency of histopathological diagnosis

Pathologists face a substantial increase in workload and complexity of histopathologic cancer diagnosis due to the advent of personalized medicine. Therefore, diagnostic protocols have to focus equally on efficiency and accuracy. In this paper we introduce ‘deep learning’ as a technique to improve the objectivity and efficiency of histopathologic slide analysis. Through two examples, prostate cancer identification in biopsy specimens and breast cancer metastasis detection in sentinel lymph nodes, we show the potential of this new methodology to reduce the workload for pathologists, while at the same time increasing objectivity of diagnoses. We found that all slides containing prostate cancer and micro- and macro-metastases of breast cancer could be identified automatically while 30–40% of the slides containing benign and normal tissue could be excluded without the use of any additional immunohistochemical markers or human intervention. We conclude that ‘deep learning’ holds great promise to improve the efficacy of prostate cancer diagnosis and breast cancer staging.

Robotic action acquisition with cognitive biases in coarse-grained state space

Some of the authors have previously proposed a cognitively inspired reinforcement learning architecture (LS-Q) that mimics cognitive biases in humans. LS-Q adaptively learns under uniform, coarse-grained state division and performs well without parameter tuning in a giant-swing robot task. However, these results were shown only in simulations. In this study, we test the validity of the LS-Q implemented in a robot in a real environment. In addition, we analyze the learning process to elucidate the mechanism by which the LS-Q adaptively learns under the partially observable environment. We argue that the LS-Q may be a versatile reinforcement learning architecture, which is, despite its simplicity, easily applicable and does not require well-prepared settings.

Automated Measurement of Cobblestone Morphology for Characterizing Stem Cell Derived Retinal Pigment Epithelial Cell Cultures

PURPOSE

Assessing the morphologic properties of cells in microscopy images is an important task to evaluate cell health, identity, and purity. Typically, subjective visual assessments are accomplished by an experienced researcher. This subjective human step makes transfer of the evaluation process from the laboratory to the cell manufacturing facility difficult and time consuming.

METHODS

Automated image analysis can provide rapid, objective measurements of cultured cells, greatly aiding manufacturing, regulatory, and research goals. Automated algorithms for classifying images based on appearance characteristics typically either extract features from the image and use those features for classification or use the images directly as input to the classification algorithm. In this study we have developed both feature and nonfeature extraction methods for automatically measuring "cobblestone" structure in human retinal pigment epithelial (RPE) cell cultures.

RESULTS

A new approach using image compression combined with a Kolmogorov complexity-based distance metric enables robust classification of microscopy images of RPE cell cultures. The automated measurements corroborate determinations made by experienced cell biologists. We have also developed an approach for using steerable wavelet filters for extracting features to characterize the individual cellular junctions.

CONCLUSIONS

Two image analysis techniques enable robust and accurate characterization of the cobblestone morphology that is indicative of viable RPE cultures for therapeutic applications.

Deep Patient: An Unsupervised Representation to Predict the Future of Patients from the Electronic Health Records

Secondary use of electronic health records (EHRs) promises to advance clinical research and better inform clinical decision making. Challenges in summarizing and representing patient data prevent widespread practice of predictive modeling using EHRs. Here we present a novel unsupervised deep feature learning method to derive a general-purpose patient representation from EHR data that facilitates clinical predictive modeling. In particular, a three-layer stack of denoising autoencoders was used to capture hierarchical regularities and dependencies in the aggregated EHRs of about 700,000 patients from the Mount Sinai data warehouse. The result is a representation we name “deep patient”. We evaluated this representation as broadly predictive of health states by assessing the probability of patients to develop various diseases. We performed evaluation using 76,214 test patients comprising 78 diseases from diverse clinical domains and temporal windows. Our results significantly outperformed those achieved using representations based on raw EHR data and alternative feature learning strategies. Prediction performance for severe diabetes, schizophrenia, and various cancers were among the top performing. These findings indicate that deep learning applied to EHRs can derive patient representations that offer improved clinical predictions, and could provide a machine learning framework for augmenting clinical decision systems.

Relaxation oscillator-realized artificial electronic neurons, their responses, and noise

A proof-of-concept relaxation oscillator-based leaky integrate-and-fire (ROLIF) neuron circuit is realized by using an amorphous chalcogenide-based threshold switch and non-ideal operational amplifier (op-amp). The proposed ROLIF neuron offers biologically plausible features such as analog-type encoding, signal amplification, unidirectional synaptic transmission, and Poisson noise. The synaptic transmission between pre- and postsynaptic neurons is achieved through a passive synapse (simple resistor). The synaptic resistor coupled to the non-ideal op-amp realizes excitatory postsynaptic potential (EPSP) evolution that evokes postsynaptic neuron spiking. In an attempt to generalize our proposed model, we theoretically examine ROLIF neuron circuits adopting different non-ideal op-amps having different gains and slew rates. The simulation results indicate the importance of gain in postsynaptic neuron spiking, irrespective of the slew rate (as long as the rate exceeds a particular value), providing the basis for the ROLIF neuron circuit design. Eventually, the behavior of a postsynaptic neuron in connection to multiple presynaptic neurons via synapses is highlighted in terms of EPSP evolution amid simultaneously incident asynchronous presynaptic spikes, which in fact reveals an important role of the random noise in spatial integration.

Discovery and Optimization of Materials Using Evolutionary Approaches

Materials science is undergoing a revolution, generating valuable new materials such as flexible solar panels, biomaterials and printable tissues, new catalysts, polymers, and porous materials with unprecedented properties. However, the number of potentially accessible materials is immense. Artificial evolutionary methods such as genetic algorithms, which explore large, complex search spaces very efficiently, can be applied to the identification and optimization of novel materials more rapidly than by physical experiments alone. Machine learning models can augment experimental measurements of materials fitness to accelerate identification of useful and novel materials in vast materials composition or property spaces. This review discusses the problems of large materials spaces, the types of evolutionary algorithms employed to identify or optimize materials, and how materials can be represented mathematically as genomes, describes fitness landscapes and mutation operators commonly employed in materials evolution, and provides a comprehensive summary of published research on the use of evolutionary methods to generate new catalysts, phosphors, and a range of other materials. The review identifies the potential for evolutionary methods to revolutionize a wide range of manufacturing, medical, and materials based industries.

Pulmonary Nodule Detection in CT Images: False Positive Reduction Using Multi-View Convolutional Networks

We propose a novel Computer-Aided Detection (CAD) system for pulmonary nodules using multi-view convolutional networks (ConvNets), for which discriminative features are automatically learnt from the training data. The network is fed with nodule candidates obtained by combining three candidate detectors specifically designed for solid, subsolid, and large nodules. For each candidate, a set of 2-D patches from differently oriented planes is extracted. The proposed architecture comprises multiple streams of 2-D ConvNets, for which the outputs are combined using a dedicated fusion method to get the final classification. Data augmentation and dropout are applied to avoid overfitting. On 888 scans of the publicly available LIDC-IDRI dataset, our method reaches high detection sensitivities of 85.4% and 90.1% at 1 and 4 false positives per scan, respectively. An additional evaluation on independent datasets from the ANODE09 challenge and DLCST is performed. We showed that the proposed multi-view ConvNets is highly suited to be used for false positive reduction of a CAD system.

Fast Convolutional Neural Network Training Using Selective Data Sampling: Application to Hemorrhage Detection in Color Fundus Images

Convolutional neural networks (CNNs) are deep learning network architectures that have pushed forward the state-of-the-art in a range of computer vision applications and are increasingly popular in medical image analysis. However, training of CNNs is time-consuming and challenging. In medical image analysis tasks, the majority of training examples are easy to classify and therefore contribute little to the CNN learning process. In this paper, we propose a method to improve and speed-up the CNN training for medical image analysis tasks by dynamically selecting misclassified negative samples during training. Training samples are heuristically sampled based on classification by the current status of the CNN. Weights are assigned to the training samples and informative samples are more likely to be included in the next CNN training iteration. We evaluated and compared our proposed method by training a CNN with (SeS) and without (NSeS) the selective sampling method. We focus on the detection of hemorrhages in color fundus images. A decreased training time from 170 epochs to 60 epochs with an increased performance-on par with two human experts-was achieved with areas under the receiver operating characteristics curve of 0.894 and 0.972 on two data sets. The SeS CNN statistically outperformed the NSeS CNN on an independent test set.

Convolutional Neural Networks for Medical Image Analysis: Full Training or Fine Tuning?

Training a deep convolutional neural network (CNN) from scratch is difficult because it requires a large amount of labeled training data and a great deal of expertise to ensure proper convergence. A promising alternative is to fine-tune a CNN that has been pre-trained using, for instance, a large set of labeled natural images. However, the substantial differences between natural and medical images may advise against such knowledge transfer. In this paper, we seek to answer the following central question in the context of medical image analysis: Can the use of pre-trained deep CNNs with sufficient fine-tuning eliminate the need for training a deep CNN from scratch? To address this question, we considered four distinct medical imaging applications in three specialties (radiology, cardiology, and gastroenterology) involving classification, detection, and segmentation from three different imaging modalities, and investigated how the performance of deep CNNs trained from scratch compared with the pre-trained CNNs fine-tuned in a layer-wise manner. Our experiments consistently demonstrated that 1) the use of a pre-trained CNN with adequate fine-tuning outperformed or, in the worst case, performed as well as a CNN trained from scratch; 2) fine-tuned CNNs were more robust to the size of training sets than CNNs trained from scratch; 3) neither shallow tuning nor deep tuning was the optimal choice for a particular application; and 4) our layer-wise fine-tuning scheme could offer a practical way to reach the best performance for the application at hand based on the amount of available data.

Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images

Among brain tumors, gliomas are the most common and aggressive, leading to a very short life expectancy in their highest grade. Thus, treatment planning is a key stage to improve the quality of life of oncological patients. Magnetic resonance imaging (MRI) is a widely used imaging technique to assess these tumors, but the large amount of data produced by MRI prevents manual segmentation in a reasonable time, limiting the use of precise quantitative measurements in the clinical practice. So, automatic and reliable segmentation methods are required; however, the large spatial and structural variability among brain tumors make automatic segmentation a challenging problem. In this paper, we propose an automatic segmentation method based on Convolutional Neural Networks (CNN), exploring small 3 ×3 kernels. The use of small kernels allows designing a deeper architecture, besides having a positive effect against overfitting, given the fewer number of weights in the network. We also investigated the use of intensity normalization as a pre-processing step, which though not common in CNN-based segmentation methods, proved together with data augmentation to be very effective for brain tumor segmentation in MRI images. Our proposal was validated in the Brain Tumor Segmentation Challenge 2013 database (BRATS 2013), obtaining simultaneously the first position for the complete, core, and enhancing regions in Dice Similarity Coefficient metric (0.88, 0.83, 0.77) for the Challenge data set. Also, it obtained the overall first position by the online evaluation platform. We also participated in the on-site BRATS 2015 Challenge using the same model, obtaining the second place, with Dice Similarity Coefficient metric of 0.78, 0.65, and 0.75 for the complete, core, and enhancing regions, respectively.

q-Space Deep Learning: Twelve-Fold Shorter and Model-Free Diffusion MRI Scans

Numerous scientific fields rely on elaborate but partly suboptimal data processing pipelines. An example is diffusion magnetic resonance imaging (diffusion MRI), a non-invasive microstructure assessment method with a prominent application in neuroimaging. Advanced diffusion models providing accurate microstructural characterization so far have required long acquisition times and thus have been inapplicable for children and adults who are uncooperative, uncomfortable, or unwell. We show that the long scan time requirements are mainly due to disadvantages of classical data processing. We demonstrate how deep learning, a group of algorithms based on recent advances in the field of artificial neural networks, can be applied to reduce diffusion MRI data processing to a single optimized step. This modification allows obtaining scalar measures from advanced models at twelve-fold reduced scan time and detecting abnormalities without using diffusion models. We set a new state of the art by estimating diffusion kurtosis measures from only 12 data points and neurite orientation dispersion and density measures from only 8 data points. This allows unprecedentedly fast and robust protocols facilitating clinical routine and demonstrates how classical data processing can be streamlined by means of deep learning.

CoinFold: a web server for protein contact prediction and contact-assisted protein folding

CoinFold (http://raptorx2.uchicago.edu/ContactMap/) is a web server for protein contact prediction and contact-assisted de novo structure prediction. CoinFold predicts contacts by integrating joint multi-family evolutionary coupling (EC) analysis and supervised machine learning. This joint EC analysis is unique in that it not only uses residue coevolution information in the target protein family, but also that in the related families which may have divergent sequences but similar folds. The supervised learning further improves contact prediction accuracy by making use of sequence profile, contact (distance) potential and other information. Finally, this server predicts tertiary structure of a sequence by feeding its predicted contacts and secondary structure to the CNS suite. Tested on the CASP and CAMEO targets, this server shows significant advantages over existing ones of similar category in both contact and tertiary structure prediction.

RaptorX-Property: a web server for protein structure property prediction

RaptorX Property (http://raptorx2.uchicago.edu/StructurePropertyPred/predict/) is a web server predicting structure property of a protein sequence without using any templates. It outperforms other servers, especially for proteins without close homologs in PDB or with very sparse sequence profile (i.e. carries little evolutionary information). This server employs a powerful in-house deep learning model DeepCNF (Deep Convolutional Neural Fields) to predict secondary structure (SS), solvent accessibility (ACC) and disorder regions (DISO). DeepCNF not only models complex sequence–structure relationship by a deep hierarchical architecture, but also interdependency between adjacent property labels. Our experimental results show that, tested on CASP10, CASP11 and the other benchmarks, this server can obtain ∼84% Q3 accuracy for 3-state SS, ∼72% Q8 accuracy for 8-state SS, ∼66% Q3 accuracy for 3-state solvent accessibility, and ∼0.89 area under the ROC curve (AUC) for disorder prediction.

Day-Ahead Prediction of Wind Speed with Deep Feature Learning

Day-ahead prediction of wind speed is a basic and key problem of large-scale wind power penetration. Many current techniques fail to satisfy practical engineering requirements because of wind speed's strong nonlinear features, influenced by many complex factors, and the general model's inability to automatically learn features. It is well recognized that wind speed varies in different patterns. In this paper, we propose a deep feature learning (DFL) approach to wind speed forecasting because of its advantages at both multi-layer feature extraction and unsupervised learning. A deep belief network (DBN) model for regression with an architecture of 144 input and 144 output nodes was constructed using a restricted Boltzmann machine (RBM). Day-ahead prediction experiments were then carried out. By comparing the experimental results, it was found that the prediction errors with respect to both size and stability of a DBN model with only three hidden layers were less than those of the other three typical approaches including support vector regression (SVR), single hidden layer neural networks (SHL-NN), and neural networks with three hidden layers (THL-NN). In addition, the DBN model can learn and obtain complex features of wind speed through its strong nonlinear mapping ability, which effectively improves its prediction precision. In addition, prediction errors are minimized when the number of DBN model's hidden layers reaches a threshold value. Above this number, it is not possible to improve the prediction accuracy by further increasing the number of hidden layers. Thus, the DBN method has a high practical value for wind speed prediction.

Progress in Fully Automated Abdominal CT Interpretation

OBJECTIVE

Automated analysis of abdominal CT has advanced markedly over just the last few years. Fully automated assessment of organs, lymph nodes, adipose tissue, muscle, bowel, spine, and tumors are some examples where tremendous progress has been made. Computer-aided detection of lesions has also improved dramatically.

CONCLUSION

This article reviews the progress and provides insights into what is in store in the near future for automated analysis for abdominal CT, ultimately leading to fully automated interpretation.

A Hybrid Model for the Computationally-Efficient Simulation of the Cerebellar Granular Layer

The aim of the present paper is to efficiently describe the membrane potential dynamics of neural populations formed by species having a high density difference in specific brain areas. We propose a hybrid model whose main ingredients are a conductance-based model (ODE system) and its continuous counterpart (PDE system) obtained through a limit process in which the number of neurons confined in a bounded region of the brain tissue is sent to infinity. Specifically, in the discrete model, each cell is described by a set of time-dependent variables, whereas in the continuum model, cells are grouped into populations that are described by a set of continuous variables. Communications between populations, which translate into interactions among the discrete and the continuous models, are the essence of the hybrid model we present here. The cerebellum and cerebellum-like structures show in their granular layer a large difference in the relative density of neuronal species making them a natural testing ground for our hybrid model. By reconstructing the ensemble activity of the cerebellar granular layer network and by comparing our results to a more realistic computational network, we demonstrate that our description of the network activity, even though it is not biophysically detailed, is still capable of reproducing salient features of neural network dynamics. Our modeling approach yields a significant computational cost reduction by increasing the simulation speed at least 270 times. The hybrid model reproduces interesting dynamics such as local microcircuit synchronization, traveling waves, center-surround, and time-windowing.

Applications in image-based profiling of perturbations

A dramatic shift has occurred in how biologists use microscopy images. Whether experiments are small-scale or high-throughput, automatically quantifying biological properties in images is now widespread. We see yet another revolution under way: a transition towards using automated image analysis to not only identify phenotypes a biologist specifically seeks to measure ('screening') but also as an unbiased and sensitive tool to capture a wide variety of subtle features of cell (or organism) state ('profiling'). Mapping similarities among samples using image-based (morphological) profiling has tremendous potential to transform drug discovery, functional genomics, and basic biological research. Applications include target identification, lead hopping, library enrichment, functionally annotating genes/alleles, and identifying small molecule modulators of gene activity and disease-specific phenotypes.

DNA context represents transcription regulation of the gene in mouse embryonic stem cells

Understanding gene regulatory information in DNA remains a significant challenge in biomedical research. This study presents a computational approach to infer gene regulatory programs from primary DNA sequences. Using DNA around transcription start sites as attributes, our model predicts gene regulation in the gene. We find that H3K27ac around TSS is an informative descriptor of the transcription program in mouse embryonic stem cells. We build a computational model inferring the cell-type-specific H3K27ac signatures in the DNA around TSS. A comparison of embryonic stem cell and liver cell-specific H3K27ac signatures in DNA shows that the H3K27ac signatures in DNA around TSS efficiently distinguish the cell-type specific H3K27ac peaks and the gene regulation. The arrangement of the H3K27ac signatures inferred from the DNA represents the transcription regulation of the gene in mESC. We show that the DNA around transcription start sites is associated with the gene regulatory program by specific interaction with H3K27ac.

Understanding deep convolutional networks

Deep convolutional networks provide state-of-the-art classifications and regressions results over many high-dimensional problems. We review their architecture, which scatters data with a cascade of linear filter weights and nonlinearities. A mathematical framework is introduced to analyse their properties. Computations of invariants involve multiscale contractions with wavelets, the linearization of hierarchical symmetries and sparse separations. Applications are discussed.

Divide and Conquer: Sub-Grouping of ASD Improves ASD Detection Based on Brain Morphometry

Low success (<60%) in autism spectrum disorder (ASD) classification using brain morphometry from the large multi-site ABIDE dataset and inconsistent findings on brain morphometric abnormalities in ASD can be attributed to the ASD heterogeneity. In this study, we show that ASD brain morphometry is highly heterogeneous, and demonstrate that the heterogeneity can be mitigated and classification improved if autism severity (AS), verbal IQ (VIQ) and age are used with morphometric features. Morphometric features from structural MRIs (sMRIs) of 734 males (ASD: 361, controls: 373) of ABIDE were derived using FreeSurfer. Applying the Random Forest classifier, an AUC of 0.61 was achieved. Adding VIQ and age to morphometric features, AUC improved to 0.68. Sub-grouping the subjects by AS, VIQ and age improved the classification with the highest AUC of 0.8 in the moderate-AS sub-group (AS = 7–8). Matching subjects on age and/or VIQ in each sub-group further improved the classification with the highest AUC of 0.92 in the low AS sub-group (AS = 4–5). AUC decreased with AS and VIQ, and was the lowest in the mid-age sub-group (13–18 years). The important features were mainly from the frontal, temporal, ventricular, right hippocampal and left amygdala regions. However, they highly varied with AS, VIQ and age. The curvature and folding index features from frontal, temporal, lingual and insular regions were dominant in younger subjects suggesting their importance for early detection. When the experiments were repeated using the Gradient Boosting classifier similar results were obtained. Our findings suggest that identifying brain biomarkers in sub-groups of ASD can yield more robust and insightful results than searching across the whole spectrum. Further, it may allow identification of sub-group specific brain biomarkers that are optimized for early detection and monitoring, increasing the utility of sMRI as an important tool for early detection of ASD.

Radiomics: a new application from established techniques

The increasing use of biomarkers in cancer have led to the concept of personalized medicine for patients. Personalized medicine provides better diagnosis and treatment options available to clinicians. Radiological imaging techniques provide an opportunity to deliver unique data on different types of tissue. However, obtaining useful information from all radiological data is challenging in the era of "big data". Recent advances in computational power and the use of genomics have generated a new area of research termed Radiomics. Radiomics is defined as the high throughput extraction of quantitative imaging features or texture (radiomics) from imaging to decode tissue pathology and creating a high dimensional data set for feature extraction. Radiomic features provide information about the gray-scale patterns, inter-pixel relationships. In addition, shape and spectral properties can be extracted within the same regions of interest on radiological images. Moreover, these features can be further used to develop computational models using advanced machine learning algorithms that may serve as a tool for personalized diagnosis and treatment guidance.

Why Neurons Have Thousands of Synapses, a Theory of Sequence Memory in Neocortex

Pyramidal neurons represent the majority of excitatory neurons in the neocortex. Each pyramidal neuron receives input from thousands of excitatory synapses that are segregated onto dendritic branches. The dendrites themselves are segregated into apical, basal, and proximal integration zones, which have different properties. It is a mystery how pyramidal neurons integrate the input from thousands of synapses, what role the different dendrites play in this integration, and what kind of network behavior this enables in cortical tissue. It has been previously proposed that non-linear properties of dendrites enable cortical neurons to recognize multiple independent patterns. In this paper we extend this idea in multiple ways. First we show that a neuron with several thousand synapses segregated on active dendrites can recognize hundreds of independent patterns of cellular activity even in the presence of large amounts of noise and pattern variation. We then propose a neuron model where patterns detected on proximal dendrites lead to action potentials, defining the classic receptive field of the neuron, and patterns detected on basal and apical dendrites act as predictions by slightly depolarizing the neuron without generating an action potential. By this mechanism, a neuron can predict its activation in hundreds of independent contexts. We then present a network model based on neurons with these properties that learns time-based sequences. The network relies on fast local inhibition to preferentially activate neurons that are slightly depolarized. Through simulation we show that the network scales well and operates robustly over a wide range of parameters as long as the network uses a sparse distributed code of cellular activations. We contrast the properties of the new network model with several other neural network models to illustrate the relative capabilities of each. We conclude that pyramidal neurons with thousands of synapses, active dendrites, and multiple integration zones create a robust and powerful sequence memory. Given the prevalence and similarity of excitatory neurons throughout the neocortex and the importance of sequence memory in inference and behavior, we propose that this form of sequence memory may be a universal property of neocortical tissue.

Applications of Deep Learning in Biomedicine

Increases in throughput and installed base of biomedical research equipment led to a massive accumulation of -omics data known to be highly variable, high-dimensional, and sourced from multiple often incompatible data platforms. While this data may be useful for biomarker identification and drug discovery, the bulk of it remains underutilized. Deep neural networks (DNNs) are efficient algorithms based on the use of compositional layers of neurons, with advantages well matched to the challenges -omics data presents. While achieving state-of-the-art results and even surpassing human accuracy in many challenging tasks, the adoption of deep learning in biomedicine has been comparatively slow. Here, we discuss key features of deep learning that may give this approach an edge over other machine learning methods. We then consider limitations and review a number of applications of deep learning in biomedical studies demonstrating proof of concept and practical utility.