Automated prediction of extubation success in extremely preterm infants: the APEX multicenter study.

BACKGROUND: Extremely preterm infants are frequently subjected to mechanical ventilation. Current prediction tools of extubation success lacks accuracy. METHODS: Multicenter study including infants with birth weight ≤1250 g undergoing their first extubation attempt. Clinical data and cardiorespiratory signals were acquired before extubation. Primary outcome was prediction of extubation success. Automated analysis of cardiorespiratory signals, development of clinical and cardiorespiratory features, and a 2-stage Clinical Decision-Balanced Random Forest classifier were used. A leave-one-out cross-validation was done. Performance was analyzed by ROC curves and determined by balanced accuracy. An exploratory analysis was performed for extubations before 7 days of age. RESULTS: A total of 241 infants were included and 44 failed (18%) extubation. The classifier had a balanced accuracy of 73% (sensitivity 70% [95% CI: 63%, 76%], specificity 75% [95% CI: 62%, 88%]). As an additional clinical-decision tool, the classifier would have led to an increase in extubation success from 82% to 93% but misclassified 60 infants who would have been successfully extubated. In infants extubated before 7 days of age, the classifier identified 16/18 failures (specificity 89%) and 73/105 infants with success (sensitivity 70%). CONCLUSIONS: Machine learning algorithms may improve a balanced prediction of extubation outcomes, but further refinement and validation is required. IMPACT: A machine learning-derived predictive model combining clinical data with automated analyses of individual cardiorespiratory signals may improve the prediction of successful extubation and identify infants at higher risk of failure with a good balanced accuracy. Such multidisciplinary approach including medicine, biomedical engineering and computer science is a step forward as current tools investigated to predict extubation outcomes lack sufficient balanced accuracy to justify their use in future trials or clinical practice. Thus, this individualized assessment can optimize patient selection for future trials of extubation readiness by decreasing exposure of low-risk infants to interventions and maximize the benefits of those at high risk.

Estimating individual treatment effect on disability progression in multiple sclerosis using deep learning.

Disability progression in multiple sclerosis remains resistant to treatment. The absence of a suitable biomarker to allow for phase 2 clinical trials presents a high barrier for drug development. We propose to enable short proof-of-concept trials by increasing statistical power using a deep-learning predictive enrichment strategy. Specifically, a multi-headed multilayer perceptron is used to estimate the conditional average treatment effect (CATE) using baseline clinical and imaging features, and patients predicted to be most responsive are preferentially randomized into a trial. Leveraging data from six randomized clinical trials (n = 3,830), we first pre-trained the model on the subset of relapsing-remitting MS patients (n = 2,520), then fine-tuned it on a subset of primary progressive MS (PPMS) patients (n = 695). In a separate held-out test set of PPMS patients randomized to anti-CD20 antibodies or placebo (n = 297), the average treatment effect was larger for the 50% (HR, 0.492; 95% CI, 0.266-0.912; p = 0.0218) and 30% (HR, 0.361; 95% CI, 0.165-0.79; p = 0.008) predicted to be most responsive, compared to 0.743 (95% CI, 0.482-1.15; p = 0.179) for the entire group. The same model could also identify responders to laquinimod in another held-out test set of PPMS patients (n = 318). Finally, we show that using this model for predictive enrichment results in important increases in power.

PhyloPGM: boosting regulatory function prediction accuracy using evolutionary information.

MOTIVATION: The computational prediction of regulatory function associated with a genomic sequence is of utter importance in -omics study, which facilitates our understanding of the underlying mechanisms underpinning the vast gene regulatory network. Prominent examples in this area include the binding prediction of transcription factors in DNA regulatory regions, and predicting RNA-protein interaction in the context of post-transcriptional gene expression. However, existing computational methods have suffered from high false-positive rates and have seldom used any evolutionary information, despite the vast amount of available orthologous data across multitudes of extant and ancestral genomes, which readily present an opportunity to improve the accuracy of existing computational methods. RESULTS: In this study, we present a novel probabilistic approach called PhyloPGM that leverages previously trained TFBS or RNA-RBP binding predictors by aggregating their predictions from various orthologous regions, in order to boost the overall prediction accuracy on human sequences. Throughout our experiments, PhyloPGM has shown significant improvement over baselines such as the sequence-based RNA-RBP binding predictor RNATracker and the sequence-based TFBS predictor that is known as FactorNet. PhyloPGM is simple in principle, easy to implement and yet, yields impressive results. AVAILABILITY AND IMPLEMENTATION: The PhyloPGM package is available at https://github.com/BlanchetteLab/PhyloPGM. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Deep learning, reinforcement learning, and world models.

Deep learning (DL) and reinforcement learning (RL) methods seem to be a part of indispensable factors to achieve human-level or super-human AI systems. On the other hand, both DL and RL have strong connections with our brain functions and with neuroscientific findings. In this review, we summarize talks and discussions in the "Deep Learning and Reinforcement Learning" session of the symposium, International Symposium on Artificial Intelligence and Brain Science. In this session, we discussed whether we can achieve comprehensive understanding of human intelligence based on the recent advances of deep learning and reinforcement learning algorithms. Speakers contributed to provide talks about their recent studies that can be key technologies to achieve human-level intelligence.

Fast reinforcement learning with generalized policy updates.

The combination of reinforcement learning with deep learning is a promising approach to tackle important sequential decision-making problems that are currently intractable. One obstacle to overcome is the amount of data needed by learning systems of this type. In this article, we propose to address this issue through a divide-and-conquer approach. We argue that complex decision problems can be naturally decomposed into multiple tasks that unfold in sequence or in parallel. By associating each task with a reward function, this problem decomposition can be seamlessly accommodated within the standard reinforcement-learning formalism. The specific way we do so is through a generalization of two fundamental operations in reinforcement learning: policy improvement and policy evaluation. The generalized version of these operations allow one to leverage the solution of some tasks to speed up the solution of others. If the reward function of a task can be well approximated as a linear combination of the reward functions of tasks previously solved, we can reduce a reinforcement-learning problem to a simpler linear regression. When this is not the case, the agent can still exploit the task solutions by using them to interact with and learn about the environment. Both strategies considerably reduce the amount of data needed to solve a reinforcement-learning problem.

Assessment of Extubation Readiness Using Spontaneous Breathing Trials in Extremely Preterm Neonates.

Importance: Spontaneous breathing trials (SBTs) are used to determine extubation readiness in extremely preterm neonates (gestational age ≤28 weeks), but these trials rely on empirical combinations of clinical events during endotracheal continuous positive airway pressure (ET-CPAP). Objectives: To describe clinical events during ET-CPAP and to assess accuracy of comprehensive clinical event combinations in predicting successful extubation compared with clinical judgment alone. Design, Setting, and Participants: This multicenter diagnostic study used data from 259 neonates seen at 5 neonatal intensive care units from the prospective Automated Prediction of Extubation Readiness (APEX) study from September 1, 2013, through August 31, 2018. Neonates with birth weight less than 1250 g who required mechanical ventilation were eligible. Neonates deemed to be ready for extubation and who underwent ET-CPAP before extubation were included. Interventions: In the APEX study, cardiorespiratory signals were recorded during 5-minute ET-CPAP, and signs of clinical instability were monitored. Main Outcomes and Measures: Four clinical events were documented during ET-CPAP: apnea requiring stimulation, presence and cumulative durations of bradycardia and desaturation, and increased supplemental oxygen. Clinical event occurrence was assessed and compared between extubation pass and fail (defined as reintubation within 7 days). An automated algorithm was developed to generate SBT definitions using all clinical event combinations and to compute diagnostic accuracies of an SBT in predicting extubation success. Results: Of 259 neonates (139 [54%] male) with a median gestational age of 26.1 weeks (interquartile range [IQR], 24.9-27.4 weeks) and median birth weight of 830 g (IQR, 690-1019 g), 147 (57%) had at least 1 clinical event during ET-CPAP. Apneas occurred in 10% (26 of 259) of neonates, bradycardias in 19% (48), desaturations in 53% (138), and increased oxygen needs in 41% (107). Neonates with successful extubation (71% [184 of 259]) had significantly fewer clinical events (51% [93 of 184] vs 72% [54 of 75], P = .002), shorter cumulative bradycardia duration (median, 0 seconds [IQR, 0 seconds] vs 0 seconds [IQR, 0-9 seconds], P < .001), shorter cumulative desaturation duration (median, 0 seconds [IQR, 0-59 seconds] vs 25 seconds [IQR, 0-90 seconds], P = .003), and less increase in oxygen (median, 0% [IQR, 0%-6%] vs 5% [0%-18%], P < .001) compared with neonates with failed extubation. In total, 41 602 SBT definitions were generated, demonstrating sensitivities of 51% to 100% (median, 96%) and specificities of 0% to 72% (median, 22%). Youden indices for all SBTs ranged from 0 to 0.32 (median, 0.17), suggesting low accuracy. The SBT with highest Youden index defined SBT pass as having no apnea (with desaturation requiring stimulation) or increase in oxygen requirements by 15% from baseline and predicted extubation success with a sensitivity of 93% and a specificity of 39%. Conclusions and Relevance: The findings suggest that extremely preterm neonates commonly show signs of clinical instability during ET-CPAP and that the accuracy of multiple clinical event combinations to define SBTs is low. Thus, SBTs may provide little added value in the assessment of extubation readiness.

Exploring uncertainty measures in deep networks for Multiple sclerosis lesion detection and segmentation.

Deep learning networks have recently been shown to outperform other segmentation methods on various public, medical-image challenge datasets, particularly on metrics focused on large pathologies. For diseases such as Multiple Sclerosis (MS), however, monitoring all the focal lesions visible on MRI sequences, even very small ones, is essential for disease staging, prognosis, and evaluating treatment efficacy. Small lesion segmentation presents significant challenges to popular deep learning models. This, coupled with their deterministic predictions, hinders their clinical adoption. Uncertainty estimates for these predictions would permit subsequent revision by clinicians. We present the first exploration of multiple uncertainty estimates based on Monte Carlo (MC) dropout (Gal and Ghahramani, 2016) in the context of deep networks for lesion detection and segmentation in medical images. Specifically, we develop a 3D MS lesion segmentation CNN, augmented to provide four different voxel-based uncertainty measures based on MC dropout. We train the network on a proprietary, large-scale, multi-site, multi-scanner, clinical MS dataset, and compute lesion-wise uncertainties by accumulating evidence from voxel-wise uncertainties within detected lesions. We analyze the performance of voxel-based segmentation and lesion-level detection by choosing operating points based on the uncertainty. Uncertainty filtering improves both voxel and lesion-wise TPR and FDR on remaining, certain predictions compared to sigmoid-based TPR/FDR curves. Small lesions and lesion-boundaries are the most uncertain regions, which is consistent with human-rater variability.

The Impact of Time Interval between Extubation and Reintubation on Death or Bronchopulmonary Dysplasia in Extremely Preterm Infants.

OBJECTIVE: To explore the relation between time to reintubation and death or bronchopulmonary dysplasia (BPD) in extremely preterm infants. STUDY DESIGN: This was a subanalysis from an ongoing multicenter observational study. Infants with birth weight ≤1250 g, requiring mechanical ventilation, and undergoing their first elective extubation were prospectively followed throughout hospitalization. Time to reintubation was defined as the time interval between first elective extubation and reintubation. Univariate and multivariate logistic regression analyses were performed to evaluate associations between time to reintubation, using different observation windows after extubation (24-hour intervals), and death/BPD (primary outcome) or BPD among survivors (secondary outcome). aORs were computed with and without the confounding effects of cumulative mechanical ventilation duration. RESULTS: Of 216 infants included for analysis, 103 (48%) were reintubated at least once after their first elective extubation. Reintubation was associated with lower gestational age/weight and greater morbidities compared with infants never reintubated. After adjusting for confounders, reintubation within observation windows ranging between 24 hours and 3 weeks postextubation was associated with increased odds of death/BPD (but not BPD among survivors), independent of the cumulative mechanical ventilation duration. Reintubation within 48 hours from extubation conferred higher risk-adjusted odds of death/BPD vs other observation windows. CONCLUSIONS: Although reintubation after elective extubation was independently associated with increased likelihood of death/BPD in extremely preterm infants, the greatest risk was attributable to reintubation within the first 48 hours postextubation. Prediction models capable of identifying the highest-risk infants may further improve outcomes.

Undersampling and Bagging of Decision Trees in the Analysis of Cardiorespiratory Behavior for the Prediction of Extubation Readiness in Extremely Preterm Infants.

Extremely preterm infants often require endotracheal intubation and mechanical ventilation during the first days of life. Due to the detrimental effects of prolonged invasive mechanical ventilation (IMV), clinicians aim to extubate infants as soon as they deem them ready.Unfortunately, existing strategies for prediction of extubation readiness vary across clinicians and institutions, and lead to high reintubation rates. We present an approach using Random Forest classifiers for the analysis of cardiorespiratory variability to predict extubation readiness. We address the issue of data imbalance by employing random undersampling of examples from the majority class before training each Decision Tree in a bag. By incorporating clinical domain knowledge, we further demonstrate that our classifier could have identified 71% of infants who failed extubation, while maintaining a success detection rate of 78%.

Patterns of reintubation in extremely preterm infants: a longitudinal cohort study.

BackgroundThe optimal approach for reporting reintubation rates in extremely preterm infants is unknown. This study aims to longitudinally describe patterns of reintubation in this population over a broad range of observation windows following extubation.MethodsTiming and reasons for reintubation following a first planned extubation were collected from infants with birth weight ≤1,250 g. An algorithm was generated to discriminate between reintubations attributable to respiratory and non-respiratory causes. Frequency and cumulative distribution curves were constructed for each category using 24 h intervals. The ability of observation windows to capture respiratory-related reintubations while limiting non-respiratory reasons was assessed using a receiver operating characteristic curve.ResultsOut of 194 infants, 91 (47%) were reintubated during hospitalization; 68% for respiratory and 32% for non-respiratory reasons. Respiratory-related reintubation rates steadily increased from 0 to 14 days post-extubation before reaching a plateau. In contrast, non-respiratory reintubations were negligible in the first post-extubation week, but became predominant after 14 days. An observation window of 7 days captured 77% of respiratory-related reintubations while only including 14% of non-respiratory cases.ConclusionReintubation patterns are highly variable and affected by the reasons for reintubation and observation window used. Ideally, reintubation rates should be reported using a cumulative distribution curve over time.

A semi-Markov chain approach to modeling respiratory patterns prior to extubation in preterm infants.

After birth, extremely preterm infants often require specialized respiratory management in the form of invasive mechanical ventilation (IMV). Protracted IMV is associated with detrimental outcomes and morbidities. Premature extubation, on the other hand, would necessitate reintubation which is risky, technically challenging and could further lead to lung injury or disease. We present an approach to modeling respiratory patterns of infants who succeeded extubation and those who required reintubation which relies on Markov models. We compare the use of traditional Markov chains to semi-Markov models which emphasize cross-pattern transitions and timing information, and to multi-chain Markov models which can concisely represent non-stationarity in respiratory behavior over time. The models we developed expose specific, unique similarities as well as vital differences between the two populations.

APEX_SCOPE: A graphical user interface for visualization of multi-modal data in inter-disciplinary studies.

In multi-disciplinary studies, different forms of data are often collected for analysis. For example, APEX, a study on the automated prediction of extubation readiness in extremely preterm infants, collects clinical parameters and cardiorespiratory signals. A variety of cardiorespiratory metrics are computed from these signals and used to assign a cardiorespiratory pattern at each time. In such a situation, exploratory analysis requires a visualization tool capable of displaying these different types of acquired and computed signals in an integrated environment. Thus, we developed APEX_SCOPE, a graphical tool for the visualization of multi-modal data comprising cardiorespiratory signals, automated cardiorespiratory metrics, automated respiratory patterns, manually classified respiratory patterns, and manual annotations by clinicians during data acquisition. This MATLAB-based application provides a means for collaborators to view combinations of signals to promote discussion, generate hypotheses and develop features.

Prediction of Extubation readiness in extremely preterm infants by the automated analysis of cardiorespiratory behavior: study protocol.

BACKGROUND: Extremely preterm infants (≤ 28 weeks gestation) commonly require endotracheal intubation and mechanical ventilation (MV) to maintain adequate oxygenation and gas exchange. Given that MV is independently associated with important adverse outcomes, efforts should be made to limit its duration. However, current methods for determining extubation readiness are inaccurate and a significant number of infants fail extubation and require reintubation, an intervention that may be associated with increased morbidities. A variety of objective measures have been proposed to better define the optimal time for extubation, but none have proven clinically useful. In a pilot study, investigators from this group have shown promising results from sophisticated, automated analyses of cardiorespiratory signals as a predictor of extubation readiness. The aim of this study is to develop an automated predictor of extubation readiness using a combination of clinical tools along with novel and automated measures of cardiorespiratory behavior, to assist clinicians in determining when extremely preterm infants are ready for extubation. METHODS: In this prospective, multicenter observational study, cardiorespiratory signals will be recorded from 250 eligible extremely preterm infants with birth weights ≤1250 g immediately prior to their first planned extubation. Automated signal analysis algorithms will compute a variety of metrics for each infant, and machine learning methods will then be used to find the optimal combination of these metrics together with clinical variables that provide the best overall prediction of extubation readiness. Using these results, investigators will develop an Automated system for Prediction of EXtubation (APEX) readiness that will integrate the software for data acquisition, signal analysis, and outcome prediction into a single application suitable for use by medical personnel in the neonatal intensive care unit. The performance of APEX will later be prospectively validated in 50 additional infants. DISCUSSION: The results of this research will provide the quantitative evidence needed to assist clinicians in determining when to extubate a preterm infant with the highest probability of success, and could produce significant improvements in extubation outcomes in this population. TRIAL REGISTRATION: Clinicaltrials.gov identifier: NCT01909947 . Registered on July 17 2013. Trial sponsor: Canadian Institutes of Health Research (CIHR).

Hierarchical Spatio-Temporal Probabilistic Graphical Model with Multiple Feature Fusion for Binary Facial Attribute Classification in Real-World Face Videos.

Recent literature shows that facial attributes, i.e., contextual facial information, can be beneficial for improving the performance of real-world applications, such as face verification, face recognition, and image search. Examples of face attributes include gender, skin color, facial hair, etc. How to robustly obtain these facial attributes (traits) is still an open problem, especially in the presence of the challenges of real-world environments: non-uniform illumination conditions, arbitrary occlusions, motion blur and background clutter. What makes this problem even more difficult is the enormous variability presented by the same subject, due to arbitrary face scales, head poses, and facial expressions. In this paper, we focus on the problem of facial trait classification in real-world face videos. We have developed a fully automatic hierarchical and probabilistic framework that models the collective set of frame class distributions and feature spatial information over a video sequence. The experiments are conducted on a large real-world face video database that we have collected, labelled and made publicly available. The proposed method is flexible enough to be applied to any facial classification problem. Experiments on a large, real-world video database McGillFaces [1] of 18,000 video frames reveal that the proposed framework outperforms alternative approaches, by up to 16.96 and 10.13%, for the facial attributes of gender and facial hair, respectively.

Automated ongoing data validation and quality control of multi-institutional studies.

This paper addresses the problem of ensuring the validity and quality of data in ongoing multi-disciplinary studies where data acquisition spans several geographical sites. It describes an automated validation and quality control procedure that requires no user supervision and monitors data acquired from different locations before analysis. The procedure is illustrated for the Automated Prediction of Extubation readiness (APEX) project in preterm infants, where acquisition of clinical and cardiorespiratory data occurs at 6 sites using different equipment and personnel. We have identified more than 40 problems with clinical information and 25 possible problems with the cardiorespiratory signals. Our validation and quality control procedure identifies these problems in an ongoing manner so that they can be timely addressed and corrected throughout this long-term collaborative study.

IMaGe: Iterative Multilevel Probabilistic Graphical Model for Detection and Segmentation of Multiple Sclerosis Lesions in Brain MRI.

In this paper, we present IMaGe, a new, iterative two-stage probabilistic graphical model for detection and segmentation of Multiple Sclerosis (MS) lesions. Our model includes two levels of Markov Random Fields (MRFs). At the bottom level, a regular grid voxel-based MRF identifies potential lesion voxels, as well as other tissue classes, using local and neighbourhood intensities and class priors. Contiguous voxels of a particular tissue type are grouped into regions. A higher, non-lattice MRF is then constructed, in which each node corresponds to a region, and edges are defined based on neighbourhood relationships between regions. The goal of this MRF is to evaluate the probability of candidate lesions, based on group intensity, texture and neighbouring regions. The inferred information is then propagated to the voxel-level MRF. This process of iterative inference between the two levels repeats as long as desired. The iterations suppress false positives and refine lesion boundaries. The framework is trained on 660 MRI volumes of MS patients enrolled in clinical trials from 174 different centres, and tested on a separate multi-centre clinical trial data set with 535 MRI volumes. All data consists of T1, T2, PD and FLAIR contrasts. In comparison to other MRF methods, such as, and a traditional MRF, IMaGe is much more sensitive (with slightly better PPV). It outperforms its nearest competitor by around 20% when detecting very small lesions (3-10 voxels). This is a significant result, as such lesions constitute around 40% of the total number of lesions.

Feature selection and oversampling in analysis of clinical data for extubation readiness in extreme preterm infants.

We present an approach for the analysis of clinical data from extremely preterm infants, in order to determine if they are ready to be removed from invasive endotracheal mechanical ventilation. The data includes over 100 clinical features, and the subject population is naturally quite small. To address this problem, we use feature selection, specifically mutual information, in order to choose a small subset of informative features. The other challenge we address is class imbalance, as there are many more babies that succeed extubation than those who fail. To handle this problem, we use SMOTE, an algorithm which creates synthetic examples of the minority class.

Correlation of clinical parameters with cardiorespiratory behavior in successfully extubated extremely preterm infants.

Extremely preterm infants (gestational age ≤ 28 weeks) often require EndoTracheal Tube-Invasive Mechanical Ventilation (ETT-IMV) to survive. Clinicians wean infants off ETT-IMV as early as possible using their judgment and clinical information. However, assessment of extubation readiness is not accurate since 20 to 40% of preterm infants fail extubation. We extended our work in automated prediction of extubation readiness by examining correlations of automated cardiorespiratory features to clinical parameters in successfully extubated infants. Only a few features, mainly those related to variability of breathing synchrony, had any consistent correlation with clinical parameters, namely gestational age, day of life at extubation, and bicarbonate. We conclude that the automated cardiorespiratory features likely provide different information additional to clinical practice.

Quantifying the determinants of outbreak detection performance through simulation and machine learning.

OBJECTIVE: To develop a probabilistic model for discovering and quantifying determinants of outbreak detection and to use the model to predict detection performance for new outbreaks. MATERIALS AND METHODS: We used an existing software platform to simulate waterborne disease outbreaks of varying duration and magnitude. The simulated data were overlaid on real data from visits to emergency department in Montreal for gastroenteritis. We analyzed the combined data using biosurveillance algorithms, varying their parameters over a wide range. We then applied structure and parameter learning algorithms to the resulting data set to build a Bayesian network model for predicting detection performance as a function of outbreak characteristics and surveillance system parameters. We evaluated the predictions of this model through 5-fold cross-validation. RESULTS: The model predicted performance metrics of commonly used outbreak detection methods with an accuracy greater than 0.80. The model also quantified the influence of different outbreak characteristics and parameters of biosurveillance algorithms on detection performance in practically relevant surveillance scenarios. In addition to identifying characteristics expected a priori to have a strong influence on detection performance, such as the alerting threshold and the peak size of the outbreak, the model suggested an important role for other algorithm features, such as adjustment for weekly patterns. CONCLUSION: We developed a model that accurately predicts how characteristics of disease outbreaks and detection methods will influence on detection. This model can be used to compare the performance of detection methods under different surveillance scenarios, to gain insight into which characteristics of outbreaks and biosurveillance algorithms drive detection performance, and to guide the configuration of surveillance systems.

Organizational principles of cloud storage to support collaborative biomedical research.

This paper describes organizational guidelines and an anonymization protocol for the management of sensitive information in interdisciplinary, multi-institutional studies with multiple collaborators. This protocol is flexible, automated, and suitable for use in cloud-based projects as well as for publication of supplementary information in journal papers. A sample implementation of the anonymization protocol is illustrated for an ongoing study dealing with Automated Prediction of EXtubation readiness (APEX).