WindowSHAP: An efficient framework for explaining time-series classifiers based on Shapley values.

Unpacking and comprehending how black-box machine learning algorithms (such as deep learning models) make decisions has been a persistent challenge for researchers and end-users. Explaining time-series predictive models is useful for clinical applications with high stakes to understand the behavior of prediction models, e.g., to determine how different variables and time points influence the clinical outcome. However, existing approaches to explain such models are frequently unique to architectures and data where the features do not have a time-varying component. In this paper, we introduce WindowSHAP, a model-agnostic framework for explaining time-series classifiers using Shapley values. We intend for WindowSHAP to mitigate the computational complexity of calculating Shapley values for long time-series data as well as improve the quality of explanations. WindowSHAP is based on partitioning a sequence into time windows. Under this framework, we present three distinct algorithms of Stationary, Sliding and Dynamic WindowSHAP, each evaluated against baseline approaches, KernelSHAP and TimeSHAP, using perturbation and sequence analyses metrics. We applied our framework to clinical time-series data from both a specialized clinical domain (Traumatic Brain Injury - TBI) as well as a broad clinical domain (critical care medicine). The experimental results demonstrate that, based on the two quantitative metrics, our framework is superior at explaining clinical time-series classifiers, while also reducing the complexity of computations. We show that for time-series data with 120 time steps (hours), merging 10 adjacent time points can reduce the CPU time of WindowSHAP by 80 % compared to KernelSHAP. We also show that our Dynamic WindowSHAP algorithm focuses more on the most important time steps and provides more understandable explanations. As a result, WindowSHAP not only accelerates the calculation of Shapley values for time-series data, but also delivers more understandable explanations with higher quality.

A self-supervised learning-based approach to clustering multivariate time-series data with missing values (SLAC-Time): An application to TBI phenotyping.

Self-supervised learning approaches provide a promising direction for clustering multivariate time-series data. However, real-world time-series data often include missing values, and the existing approaches require imputing missing values before clustering, which may cause extensive computations and noise and result in invalid interpretations. To address these challenges, we present a Self-supervised Learning-based Approach to Clustering multivariate Time-series data with missing values (SLAC-Time). SLAC-Time is a Transformer-based clustering method that uses time-series forecasting as a proxy task for leveraging unlabeled data and learning more robust time-series representations. This method jointly learns the neural network parameters and the cluster assignments of the learned representations. It iteratively clusters the learned representations with the K-means method and then utilizes the subsequent cluster assignments as pseudo-labels to update the model parameters. To evaluate our proposed approach, we applied it to clustering and phenotyping Traumatic Brain Injury (TBI) patients in the Transforming Research and Clinical Knowledge in Traumatic Brain Injury (TRACK-TBI) study. Clinical data associated with TBI patients are often measured over time and represented as time-series variables characterized by missing values and irregular time intervals. Our experiments demonstrate that SLAC-Time outperforms the baseline K-means clustering algorithm in terms of silhouette coefficient, Calinski Harabasz index, Dunn index, and Davies Bouldin index. We identified three TBI phenotypes that are distinct from one another in terms of clinically significant variables as well as clinical outcomes, including the Extended Glasgow Outcome Scale (GOSE) score, Intensive Care Unit (ICU) length of stay, and mortality rate. The experiments show that the TBI phenotypes identified by SLAC-Time can be potentially used for developing targeted clinical trials and therapeutic strategies.

Femoral vascular conductance and peroneal muscle sympathetic nerve activity responses to acute epidural spinal cord stimulation in humans.

NEW FINDINGS: What is the central question of this research? Does acute spinal cord stimulation increase vascular conductance and decrease muscle sympathetic nerve activity in the lower limbs of humans? What is the main finding and its importance? Acute spinal cord stimulation led to a rapid rise in femoral vascular conductance, and peroneal muscle sympathetic nerve activity demonstrated a delayed reduction that was not associated with the initial increase in femoral vascular conductance. These findings suggest that neural mechanisms in addition to attenuated muscle sympathetic nerve activity might be involved in the initial increase in femoral vascular conductance during acute spinal cord stimulation. ABSTRACT: Clinical cases have indicated an increase in peripheral blood flow after continuous epidural spinal cord stimulation (SCS) and that reduced muscle sympathetic nerve activity (MSNA) might be a potential mechanism. However, no studies in humans have directly examined the effects of acute SCS (<60 min) on vascular conductance and MSNA. In study 1, we tested the hypothesis that acute SCS (<60 min) of the thoracic spine would lead to increased common femoral vascular conductance, but not brachial vascular conductance, in 11 patients who previously underwent surgical SCS implantation for management of neuropathic pain. Throughout 60 min of SCS, common femoral artery conductance was elevated and significantly different from brachial artery conductance [in millilitres per minute: 15 min, change (Δ) 26 ± 37 versus Δ-2 ± 19%; 30 min, Δ28 ± 45 versus Δ0 ± 26%; 45 min, Δ48 ± 43 versus Δ2 ± 21%; 60 min, Δ36 ± 61 versus Δ1 ± 24%; and 15 min post-SCS, Δ51 ± 64 versus Δ6 ± 33%; P = 0.013]. A similar examination in a patient with cervical SCS revealed minimal changes in vascular conductance. In study 2, we examined whether acute SCS reduces peroneal MSNA in a subset of SCS patients (n = 5). The MSNA burst incidence in response to acute SCS gradually declined and was significantly reduced at 45 and 60 min of SCS (in bursts per 100 heart beats: 15 min, Δ-1 ± 12%; 30 min, Δ-14 ± 12%; 45 min, Δ-19 ± 16%; 60 min, Δ-24 ± 18%; and 15 min post-SCS: Δ-11 ± 7%; P = 0.015). These data demonstrate that acute SCS rapidly increases femoral vascular conductance and reduces peroneal MSNA. The gradual reduction in peroneal MSNA observed during acute SCS suggests that neural mechanisms in addition to attenuated MSNA might be involved in the acute increase in femoral vascular conductance.

Intrathecal Therapeutics: Device Design, Access Methods, and Complication Mitigation.

INTRODUCTION: The intrathecal space remains underutilized for diagnostic testing, invasive monitoring or as a pipeline for the delivery of neurological therapeutic agents and devices. The latter including drug infusions, implants for electrical modulation, and a means for maintaining the physiologic pressure column. The reasons for this are many but include unfamiliarity with the central nervous system and the historical risks that continue to overshadow the low complication rates in modern clinical series. MATERIALS AND METHODS: Our intent in this review is to explore the access devices currently on the market, assess the risk associated with breaching the intrathecal space, and propose a research model for bringing to patients the next generation of intrathecal hardware. For this purpose, we reviewed both historical and contemporary literature that pertains to the access devices and catheters intended for both temporary and permanent implantation and the complications thereof. RESULTS: There are few devices that are currently marketed in the United States or Europe for intrathecal use. Most hew to a relatively fixed design pattern predicated on the dimensions and properties of the thecal sac. All are typically composed of soft silicone, and employ a Tuohy needle for access despite design limitations. In general, these catheters are engineered for durability, ease of use, and regional deployment. Devices on the market with steerability or targeted intrathecal fixation are not yet available. Complications, once a legitimate concern, are now quite rare when recommended techniques are followed. CONCLUSIONS: Over the next decade, advances in intrathecal catheter design, access techniques, imaging, and greater understanding of the spinal cord neurophysiology will usher in an era where the intrathecal space is recognized as a highly valued diagnostic and therapeutic target. We anticipate that this will occur in several concurrent phases, each with the potential to accelerate the growth of the others.

Novel technique for trialing peripheral nerve stimulation: ultrasonography-guided StimuCath trial.

OBJECTIVE Peripheral nerve stimulation (PNS) has been used for the treatment of neuropathic pain for many decades. Despite the specific indications for PNS, clinicians often have difficulty screening for candidates likely to have a good or fair outcome. Given the expense of a permanent implant, most insurance companies will not pay for the implant without a successful PNS trial. And since PNS has only recently been approved by the US Food and Drug Administration, many insurance companies will not pay for a conventional trial of PNS. The objective of this study is to describe a short low-cost method for trialing and screening patients for peripheral nerve stimulator implantation. Additionally, this study demonstrates the long-term efficacy of PNS in the treatment of chronic neuropathic pain and the relative effectiveness of this novel screening method. METHODS The records of all patients who had undergone trialing and implantation of a PNS system for chronic refractory pain at the authors' institution over a 1-year period (August 1, 2012-July 31, 2013) were examined in this retrospective case series. The search revealed 17 patients, 13 who had undergone a novel in-office ultrasonography-guided StimuCath screening technique and 4 who had undergone a traditional week-long screening procedure. All 17 patients experienced a successful PNS trial and proceeded to permanent PNS system implantation. Patients were followed up for a mean duration of 3.0 years. Visual analog scale (VAS) pain scores were used to assess pain relief in the short-term (< 6 weeks), at 1 year, and at the last follow-up. Final outcome was also characterized as good, fair, poor, or bad. RESULTS Of these 17 patients, 10 were still using their stimulator at the last follow-up, with 8 of them obtaining good relief (classified as ≥ 50% pain relief, with an average 81% reduction in the VAS score) and 2 patients attaining fair relief (< 50% relief but still using stimulation therapy). Among the remaining 7 patients, the stimulator had been explanted in 4 and there had been no relief in 3. Excluding explanted cases, follow-up ranged from 14 to 46 months, with an average of 36 months. Patients with good or fair relief had experienced pain prior to implantation for an average of 5.1 years (range 1.8-15.2 years). A longer duration of pain trended toward a poorer outcome (bad outcome 7.6 years vs good outcome 4.1 years, p = 0.03). Seven (54%) of the 13 patients with the shorter trial experienced a good or fair outcome with an average 79% reduction in the VAS score; however, all 4 of the bad outcome cases came from this group. Three (75%) of the 4 patients with the longer trial experienced a good or fair outcome at the last follow-up, with an average 54% reduction in the VAS score. There was no difference between the trialing methods and the proportion of favorable (good or fair) outcomes (p = 0.71). CONCLUSIONS Short, ultrasonography-guided StimuCath trials were feasible in screening patients for permanent implantation of PNS, with efficacy similar to the traditional week-long screening noted at the 3-year follow-up.

Spinal Cord Stimulation for Spasticity: Historical Approaches, Current Status, and Future Directions.

INTRODUCTION: Millions of people worldwide suffer with spasticity related to irreversible damage to the brain or spinal cord. Typical antecedent events include stroke, traumatic brain injury, and spinal cord injury, although insidious onset is also common. Regardless of the cause, the resulting spasticity leads to years of disability and reduced quality of life. Many treatments are available to manage spasticity; yet each is fraught with drawbacks including incomplete response, high cost, limited duration, dose-limiting side effects, and periodic maintenance. Spinal cord stimulation (SCS), a once promising therapy for spasticity, has largely been relegated to permanent experimental status. METHODS: In this review, our goal is to document and critique the history and assess the development of SCS as a treatment of lower limb spasticity. By incorporating recent discoveries with the insights gained from the early pioneers in this field, we intend to lay the groundwork needed to propose testable hypotheses for future studies. RESULTS: SCS has been tested in over 25 different conditions since a potentially beneficial effect was first reported in 1973. However, the lack of a fully formed understanding of the pathophysiology of spasticity, archaic study methodology, and the early technological limitations of implantable hardware limit the validity of many studies. SCS offers a measure of control for spasticity that cannot be duplicated with other interventions. CONCLUSIONS: With improved energy-source miniaturization, tailored control algorithms, novel implant design, and a clearer picture of the pathophysiology of spasticity, we are poised to reintroduce and test SCS in this population.

An individualized super-Gaussian single microphone Speech Enhancement for hearing aid users with smartphone as an assistive device.

In this letter, we derive a new super Gaussian Joint Maximum a Posteriori (SGJMAP) based single microphone speech enhancement gain function. The developed Speech Enhancement method is implemented on a smartphone, and this arrangement functions as an assistive device to hearing aids. We introduce a "tradeoff" parameter in the derived gain function that allows the smartphone user to customize their listening preference, by controlling the amount of noise suppression and speech distortion in real-time based on their level of hearing comfort perceived in noisy real world acoustic environment. Objective quality and intelligibility measures show the effectiveness of the proposed method in comparison to benchmark techniques considered in this paper. Subjective results reflect the usefulness of the developed Speech Enhancement application in real-world noisy conditions at signal to noise ratio levels of -5 dB, 0 dB and 5 dB.

Charcot-marie-tooth disease type 1X in women: Electrodiagnostic findings.

INTRODUCTION: Symptoms and signs in women with Charcot-Marie-Tooth disease type 1X (CMT1X) are often milder from those in men, but the available electrophysiologic evidence regarding CMT1X in women has been characterized in some patients as non-uniform or asymmetric. METHODS: We retrospectively reviewed electrodiagnostic findings from 45 women and 31 men with CMT1X. RESULTS: Motor nerve conduction parameters in CMT1X women were less abnormal (P < 0.05), and a wider range of motor conduction velocities (CVs) were seen in women (P < 0.001) compared with men. In women, nerve conduction studies showed lack of conduction block without temporal dispersion. Motor CVs were more frequently in the normal range in women compared with men. There was no significant relationship to age of presentation and motor CV or compound muscle action potential in women. CONCLUSION: NCS parameters in CMT1X women did not demonstrate features suggestive of an acquired demyelinating neuropathy. Muscle Nerve, 2016 Muscle Nerve 54: -, 2016 Muscle Nerve 54: 728-732, 2016.

Prolongation of F-wave minimal latency: a sensitive predictor of polyneuropathy.

INTRODUCTION: To evaluate the sensitivity of F-wave minimal latencies, we compared F-waves with motor and sensory nerve conduction studies (MNCS and SNCS) in patients with peripheral neuropathy. METHODS: A retrospective chart review conducted in 484 patients confirmed the clinical evidence of a polyneuropathy, and studies of F-wave minimal latencies as well as MNCS and SNCS in each patient. RESULTS: Overall rate of abnormality reached 469/484 (96.9%) for F-wave minimal latencies as compared to 374/484 (77%) for nerve conduction studies ( p < 0.0001). Nerve-specific abnormalities of F-waves showed 290/354 (82%), 140/171 (82%), 367/398 (92%) and 357/376 (95%) for median, ulnar, peroneal and tibial nerves, respectively. Corresponding values for MNCS consisted of 108/354 (31%), 29/171 (17%), 258/398 (65%) and 189/376 (50%) (all p < 0.0001). In contrast, SNCS revealed abnormalities in 120/333 (36%), 60/159 (38%) and 266/474 (56%) of median, ulnar and sural nerves. CONCLUSION: F-wave minimal latencies serve as the best predictor of polyneuropathy followed by SNCS and then MNCS.

Transfer Learning for Class Imbalance Problems with Inadequate Data.

A fundamental problem in data mining is to effectively build robust classifiers in the presence of skewed data distributions. Class imbalance classifiers are trained specifically for skewed distribution datasets. Existing methods assume an ample supply of training examples as a fundamental prerequisite for constructing an effective classifier. However, when sufficient data is not readily available, the development of a representative classification algorithm becomes even more difficult due to the unequal distribution between classes. We provide a unified framework that will potentially take advantage of auxiliary data using a transfer learning mechanism and simultaneously build a robust classifier to tackle this imbalance issue in the presence of few training samples in a particular target domain of interest. Transfer learning methods use auxiliary data to augment learning when training examples are not sufficient and in this paper we will develop a method that is optimized to simultaneously augment the training data and induce balance into skewed datasets. We propose a novel boosting based instance-transfer classifier with a label-dependent update mechanism that simultaneously compensates for class imbalance and incorporates samples from an auxiliary domain to improve classification. We provide theoretical and empirical validation of our method and apply to healthcare and text classification applications.

A Distributed Ensemble Approach for Mining Healthcare Data under Privacy Constraints.

In recent years, electronic health records (EHRs) have been widely adapted at many healthcare facilities in an attempt to improve the quality of patient care and increase the productivity and efficiency of healthcare delivery. These EHRs can accurately diagnose diseases if utilized appropriately. While the EHRs can potentially resolve many of the existing problems associated with disease diagnosis, one of the main obstacles in effectively using them is the patient privacy and sensitivity of the medical information available in the EHR. Due to these concerns, even if the EHRs are available for storage and retrieval purposes, sharing of the patient records between different healthcare facilities has become a major concern and has hampered some of the effective advantages of using EHRs. Due to this lack of data sharing, most of the facilities aim at building clinical decision support systems using limited amount of patient data from their own EHR systems to provide important diagnosis related decisions. It becomes quite infeasible for a newly established healthcare facility to build a robust decision making system due to the lack of sufficient patient records. However, to make effective decisions from clinical data, it is indispensable to have large amounts of data to train the decision models. In this regard, there are conflicting objectives of preserving patient privacy and having sufficient data for modeling and decision making. To handle such disparate goals, we develop two adaptive distributed privacy-preserving algorithms based on a distributed ensemble strategy. The basic idea of our approach is to build an elegant model for each participating facility to accurately learn the data distribution, and then can transfer the useful healthcare knowledge acquired on their data from these participators in the form of their own decision models without revealing and sharing the patient-level sensitive data, thus protecting patient privacy. We demonstrate that our approach can successfully build accurate and robust prediction models, under privacy constraints, using the healthcare data collected from different geographical locations. We demonstrate the performance of our method using the Type-2 diabetes EHRs accumulated from multiple sources from all fifty states in the U.S. Our method was evaluated on diagnosing diabetes in the presence of insufficient number of patient records from certain regions without revealing the actual patient data from other regions. Using the proposed approach, we also discovered the important biomarkers, both universal and region-specific, and validated the selected biomarkers using the biomedical literature.

Combined common peroneal and tibial nerve injury after knee dislocation: one injury or two? An MRI-clinical correlation.

OBJECT Knee dislocations are often accompanied by stretch injuries to the common peroneal nerve (CPN). A small subset of these injuries also affect the tibial nerve. The mechanism of this combined pattern could be a single longitudinal stretch injury of the CPN extending to the sciatic bifurcation (and tibial division) or separate injuries of both the CPN and tibial nerve, either at the level of the tibiofemoral joint or distally at the soleal sling and fibular neck. The authors reviewed cases involving patients with knee dislocations with CPN and tibial nerve injuries to determine the localization of the combined injury and correlation between degree of MRI appearance and clinical severity of nerve injury. METHODS Three groups of cases were reviewed. Group 1 consisted of knee dislocations with clinical evidence of nerve injury (n = 28, including 19 cases of complete CPN injury); Group 2 consisted of knee dislocations without clinical evidence of nerve injury (n = 19); and Group 3 consisted of cases of minor knee trauma but without knee dislocation (n = 14). All patients had an MRI study of the knee performed within 3 months of injury. MRI appearance of tibial and common peroneal nerve injury was scored by 2 independent radiologists in 3 zones (Zone I, sciatic bifurcation; Zone II, knee joint; and Zone III, soleal sling and fibular neck) on a severity scale of 1-4. Injury signal was scored as diffuse or focal for each nerve in each of the 3 zones. A clinical score was also calculated based on Medical Research Council scores for strength in the tibial and peroneal nerve distributions, combined with electrophysiological data, when available, and correlated with the MRI injury score. RESULTS Nearly all of the nerve segments visualized in Groups 1 and 2 demonstrated some degree of injury on MRI (95%), compared with 12% of nerve segments in Group 3. MRI nerve injury scores were significantly more severe in Group 1 relative to Group 2 (2.06 vs 1.24, p < 0.001) and Group 2 relative to Group 3 (1.24 vs 0.13, p < 0.001). In both groups of patients with knee dislocations (Groups 1 and 2), the MRI nerve injury score was significantly higher for CPN than tibial nerve (2.72 vs 1.40 for Group 1, p < 0.001; 1.39 vs 1.09 for Group 2, p < 0.05). The clinical injury score had a significantly strong correlation with the MRI injury score for the CPN (r = 0.75, p < 0.001), but not for the tibial nerve (r = 0.07, p = 0.83). CONCLUSIONS MRI is highly sensitive in detecting subclinical nerve injury. In knee dislocation, clinical tibial nerve injury is always associated with simultaneous CPN injury, but tibial nerve function is never worse than peroneal nerve function. The point of maximum injury can occur in any of 3 zones.

Efficient Mining of Discriminative Co-clusters from Gene Expression Data.

Discriminative models are used to analyze the differences between two classes and to identify class-specific patterns. Most of the existing discriminative models depend on using the entire feature space to compute the discriminative patterns for each class. Co-clustering has been proposed to capture the patterns that are correlated in a subset of features, but it cannot handle discriminative patterns in labeled datasets. In certain biological applications such as gene expression analysis, it is critical to consider the discriminative patterns that are correlated only in a subset of the feature space. The objective of this paper is two-fold: first, it presents an algorithm to efficiently find arbitrarily positioned co-clusters from complex data. Second, it extends this co-clustering algorithm to discover discriminative co-clusters by incorporating the class information into the co-cluster search process. In addition, we also characterize the discriminative co-clusters and propose three novel measures that can be used to evaluate the performance of any discriminative subspace pattern mining algorithm. We evaluated the proposed algorithms on several synthetic and real gene expression datasets, and our experimental results showed that the proposed algorithms outperformed several existing algorithms available in the literature.

Discovery of generalizable TBI phenotypes using multivariate time-series clustering.

Traumatic Brain Injury (TBI) presents a broad spectrum of clinical presentations and outcomes due to its inherent heterogeneity, leading to diverse recovery trajectories and varied therapeutic responses. While many studies have delved into TBI phenotyping for distinct patient populations, identifying TBI phenotypes that consistently generalize across various settings and populations remains a critical research gap. Our research addresses this by employing multivariate time-series clustering to unveil TBI's dynamic intricates. Utilizing a self-supervised learning-based approach to clustering multivariate time-Series data with missing values (SLAC-Time), we analyzed both the research-centric TRACK-TBI and the real-world MIMIC-IV datasets. Remarkably, the optimal hyperparameters of SLAC-Time and the ideal number of clusters remained consistent across these datasets, underscoring SLAC-Time's stability across heterogeneous datasets. Our analysis revealed three generalizable TBI phenotypes (α, ß, and γ), each exhibiting distinct non-temporal features during emergency department visits, and temporal feature profiles throughout ICU stays. Specifically, phenotype α represents mild TBI with a remarkably consistent clinical presentation. In contrast, phenotype ß signifies severe TBI with diverse clinical manifestations, and phenotype γ represents a moderate TBI profile in terms of severity and clinical diversity. Age is a significant determinant of TBI outcomes, with older cohorts recording higher mortality rates. Importantly, while certain features varied by age, the core characteristics of TBI manifestations tied to each phenotype remain consistent across diverse populations.

Discovery of Generalizable TBI Phenotypes Using Multivariate Time-Series Clustering.

Traumatic Brain Injury (TBI) presents a broad spectrum of clinical presentations and outcomes due to its inherent heterogeneity, leading to diverse recovery trajectories and varied therapeutic responses. While many studies have delved into TBI phenotyping for distinct patient populations, identifying TBI phenotypes that consistently generalize across various settings and populations remains a critical research gap. Our research addresses this by employing multivariate time-series clustering to unveil TBI's dynamic intricates. Utilizing a self-supervised learning-based approach to clustering multivariate time-Series data with missing values (SLAC-Time), we analyzed both the research-centric TRACK-TBI and the real-world MIMIC-IV datasets. Remarkably, the optimal hyperparameters of SLAC-Time and the ideal number of clusters remained consistent across these datasets, underscoring SLAC-Time's stability across heterogeneous datasets. Our analysis revealed three generalizable TBI phenotypes (α, ß, and γ), each exhibiting distinct non-temporal features during emergency department visits, and temporal feature profiles throughout ICU stays. Specifically, phenotype α represents mild TBI with a remarkably consistent clinical presentation. In contrast, phenotype ß signifies severe TBI with diverse clinical manifestations, and phenotype γ represents a moderate TBI profile in terms of severity and clinical diversity. Age is a significant determinant of TBI outcomes, with older cohorts recording higher mortality rates. Importantly, while certain features varied by age, the core characteristics of TBI manifestations tied to each phenotype remain consistent across diverse populations.

Charcot arthropathy of the shoulder associated with typical and atypical findings.

We present a patient with a relatively rare condition: Charcot joint of the shoulder, with a rare complication, the first known example of combined neurovascular compression in this location. A 49-year-old man presented with neuropathic arthropathy of the shoulder caused by syringomyelia from a Chiari I malformation, leading to compression of both the brachial plexus and the axillary vein by mass effect from the synovitis. The brachial plexopathy resolved with surgical decompression and synovectomy, and the syringomyelia stabilized after Chiari decompression. A large acromioclavicular joint synovial cyst developed as a late complication, which was treated nonoperatively. Understanding neuropathic arthropathy can explain the spectrum of interrelated typical and atypical features in this case over long-term follow-up.

Self-Supervised Graph Learning With Hyperbolic Embedding for Temporal Health Event Prediction.

Electronic health records (EHRs) have been heavily used in modern healthcare systems for recording patients' admission information to health facilities. Many data-driven approaches employ temporal features in EHR for predicting specific diseases, readmission times, and diagnoses of patients. However, most existing predictive models cannot fully utilize EHR data, due to an inherent lack of labels in supervised training for some temporal events. Moreover, it is hard for the existing methods to simultaneously provide generic and personalized interpretability. To address these challenges, we propose Sherbet, a self-supervised graph learning framework with hyperbolic embeddings for temporal health event prediction. We first propose a hyperbolic embedding method with information flow to pretrain medical code representations in a hierarchical structure. We incorporate these pretrained representations into a graph neural network (GNN) to detect disease complications and design a multilevel attention method to compute the contributions of particular diseases and admissions, thus enhancing personalized interpretability. We present a new hierarchy-enhanced historical prediction proxy task in our self-supervised learning framework to fully utilize EHR data and exploit medical domain knowledge. We conduct a comprehensive set of experiments on widely used publicly available EHR datasets to verify the effectiveness of our model. Our results demonstrate the proposed model's strengths in both predictive tasks and interpretable abilities.

Identifying TBI Physiological States by Clustering Multivariate Clinical Time-Series Data.

Determining clinically relevant physiological states from multivariate time-series data with missing values is essential for providing appropriate treatment for acute conditions such as Traumatic Brain Injury (TBI), respiratory failure, and heart failure. Utilizing non-temporal clustering or data imputation and aggregation techniques may lead to loss of valuable information and biased analyses. In our study, we apply the SLAC-Time algorithm, an innovative self-supervision-based approach that maintains data integrity by avoiding imputation or aggregation, offering a more useful representation of acute patient states. By using SLAC-Time to cluster data in a large research dataset, we identified three distinct TBI physiological states and their specific feature profiles. We employed various clustering evaluation metrics and incorporated input from a clinical domain expert to validate and interpret the identified physiological states. Further, we discovered how specific clinical events and interventions can influence patient states and state transitions.

An Empirical Comparison of Explainable Artificial Intelligence Methods for Clinical Data: A Case Study on Traumatic Brain Injury.

A longstanding challenge surrounding deep learning algorithms is unpacking and understanding how they make their decisions. Explainable Artificial Intelligence (XAI) offers methods to provide explanations of internal functions of algorithms and reasons behind their decisions in ways that are interpretable and understandable to human users. . Numerous XAI approaches have been developed thus far, and a comparative analysis of these strategies seems necessary to discern their relevance to clinical prediction models. To this end, we first implemented two prediction models for short- and long-term outcomes of traumatic brain injury (TBI) utilizing structured tabular as well as time-series physiologic data, respectively. Six different interpretation techniques were used to describe both prediction models at the local and global levels. We then performed a critical analysis of merits and drawbacks of each strategy, highlighting the implications for researchers who are interested in applying these methodologies. The implemented methods were compared to one another in terms of several XAI characteristics such as understandability, fidelity, and stability. Our findings show that SHAP is the most stable with the highest fidelity but falls short of understandability. Anchors, on the other hand, is the most understandable approach, but it is only applicable to tabular data and not time series data.

Preparing for the next pandemic via transfer learning from existing diseases with hierarchical multi-modal BERT: a study on COVID-19 outcome prediction.

Developing prediction models for emerging infectious diseases from relatively small numbers of cases is a critical need for improving pandemic preparedness. Using COVID-19 as an exemplar, we propose a transfer learning methodology for developing predictive models from multi-modal electronic healthcare records by leveraging information from more prevalent diseases with shared clinical characteristics. Our novel hierarchical, multi-modal model ([Formula: see text]) integrates baseline risk factors from the natural language processing of clinical notes at admission, time-series measurements of biomarkers obtained from laboratory tests, and discrete diagnostic, procedure and drug codes. We demonstrate the alignment of [Formula: see text]'s predictions with well-established clinical knowledge about COVID-19 through univariate and multivariate risk factor driven sub-cohort analysis. [Formula: see text]'s superior performance over state-of-the-art methods shows that leveraging patient data across modalities and transferring prior knowledge from similar disorders is critical for accurate prediction of patient outcomes, and this approach may serve as an important tool in the early response to future pandemics.