Interpretable (not just posthoc-explainable) medical claims modeling for discharge placement to reduce preventable all-cause readmissions or death.

We developed an inherently interpretable multilevel Bayesian framework for representing variation in regression coefficients that mimics the piecewise linearity of ReLU-activated deep neural networks. We used the framework to formulate a survival model for using medical claims to predict hospital readmission and death that focuses on discharge placement, adjusting for confounding in estimating causal local average treatment effects. We trained the model on a 5% sample of Medicare beneficiaries from 2008 and 2011, based on their 2009-2011 inpatient episodes (approximately 1.2 million), and then tested the model on 2012 episodes (approximately 400 thousand). The model scored an out-of-sample AUROC of approximately 0.75 on predicting all-cause readmissions-defined using official Centers for Medicare and Medicaid Services (CMS) methodology-or death within 30-days of discharge, being competitive against XGBoost and a Bayesian deep neural network, demonstrating that one need-not sacrifice interpretability for accuracy. Crucially, as a regression model, it provides what blackboxes cannot-its exact gold-standard global interpretation, explicitly defining how the model performs its internal "reasoning" for mapping the input data features to predictions. In doing so, we identify relative risk factors and quantify the effect of discharge placement. We also show that the posthoc explainer SHAP provides explanations that are inconsistent with the ground truth model reasoning that our model readily admits.

Removing ballistocardiogram (BCG) artifact from full-scalp EEG acquired inside the MR scanner with Orthogonal Matching Pursuit (OMP).

Ballistocardiogram (BCG) artifact remains a major challenge that renders electroencephalographic (EEG) signals hard to interpret in simultaneous EEG and functional MRI (fMRI) data acquisition. Here, we propose an integrated learning and inference approach that takes advantage of a commercial high-density EEG cap, to estimate the BCG contribution in noisy EEG recordings from inside the MR scanner. To estimate reliably the full-scalp BCG artifacts, a near-optimal subset (20 out of 256) of channels first was identified using a modified recording setup. In subsequent recordings inside the MR scanner, BCG-only signal from this subset of channels was used to generate continuous estimates of the full-scalp BCG artifacts via inference, from which the intended EEG signal was recovered. The reconstruction of the EEG was performed with both a direct subtraction and an optimization scheme. We evaluated the performance on both synthetic and real contaminated recordings, and compared it to the benchmark Optimal Basis Set (OBS) method. In the challenging non-event-related-potential (non-ERP) EEG studies, our reconstruction can yield more than fourteen-fold improvement in reducing the normalized RMS error of EEG signals, compared to OBS.

Separation and reconstruction of BCG and EEG signals during continuous EEG and fMRI recordings.

Despite considerable effort to remove it, the ballistocardiogram (BCG) remains a major artifact in electroencephalographic data (EEG) acquired inside magnetic resonance imaging (MRI) scanners, particularly in continuous (as opposed to event-related) recordings. In this study, we have developed a new Direct Recording Prior Encoding (DRPE) method to extract and separate the BCG and EEG components from contaminated signals, and have demonstrated its performance by comparing it quantitatively to the popular Optimal Basis Set (OBS) method. Our modified recording configuration allows us to obtain representative bases of the BCG- and EEG-only signals. Further, we have developed an optimization-based reconstruction approach to maximally incorporate prior knowledge of the BCG/EEG subspaces, and of the signal characteristics within them. Both OBS and DRPE methods were tested with experimental data, and compared quantitatively using cross-validation. In the challenging continuous EEG studies, DRPE outperforms the OBS method by nearly sevenfold in separating the continuous BCG and EEG signals.

Mechanisms for interferon-α-induced depression and neural stem cell dysfunction.

New neurons generated by the neural stem cells (NSCs) in the adult hippocampus play an important role in emotional regulation and respond to the action of antidepressants. Depression is a common and serious side effect of interferon-α (IFN-α), which limits its use as an antiviral and antitumor drug. However, the mechanism(s) underlying IFN-induced depression are largely unknown. Using a comprehensive battery of behavioral tests, we found that mice subjected to IFN-α treatment exhibited a depression-like phenotype. IFN-α directly suppressed NSC proliferation, resulting in the reduced generation of new neurons. Brain-specific mouse knockout of the IFN-α receptor prevented IFN-α-induced depressive behavioral phenotypes and the inhibition of neurogenesis, suggesting that IFN-α suppresses hippocampal neurogenesis and induces depression via its receptor in the brain. These findings provide insight for understanding the neuropathology underlying IFN-α-induced depression and for developing new strategies for the prevention and treatment of IFN-α-induced depressive effects.

Prognostic impact of microRNA-145 down-regulation in adult T-cell leukemia/lymphoma.

Adult T-cell leukemia/lymphoma (ATL) is a highly aggressive tumor caused by human T-cell leukemia virus type 1. MicroRNAs (miRNAs) are closely involved in the development and progression of various tumors. Here we investigated the dysregulation of miRNAs in ATL and its clinical significance. Studies using miRNA arrays and subsequent real-time reverse transcription polymerase chain reaction showed that, in the 9 ATL cell lines examined, 1 miRNA was consistently up-regulated, whereas another 3 were consistently down-regulated, compared with normal CD4-positive lymphocytes. Next, we analyzed the prognostic impact of these 4 miRNAs in patients with aggressive-type ATL (n = 40). Of the 4 dysregulated miRNAs selected, 3 (miR-130b higher expression, miR-145 lower expression, and miR-223 lower expression) were significantly associated with a worsened overall patient survival. We found that expressions of these 3 miRNAs were correlated with each other. To clarify which of the 3 had the most significant impact on overall survival, we performed a multivariate prognostic analysis that included these 3 miRNAs, and only miR-145 lower expression was selected as an independent risk factor (P = .0005). When overexpressed in an ATL cell line in vitro, miR-145 specifically inhibited tumor cell growth. In conclusion, our study suggests that miR-145 down-regulation provides a growth advantage in ATL and is highly associated with a worsened prognosis for patients with ALT. Hence, miR-145 may be a useful prognostic marker and a potential therapeutic target for ATL.

Genetic and functional studies implicate synaptic overgrowth and ring gland cAMP/PKA signaling defects in the Drosophila melanogaster neurofibromatosis-1 growth deficiency.

Neurofibromatosis type 1 (NF1), a genetic disease that affects 1 in 3,000, is caused by loss of a large evolutionary conserved protein that serves as a GTPase Activating Protein (GAP) for Ras. Among Drosophila melanogaster Nf1 (dNf1) null mutant phenotypes, learning/memory deficits and reduced overall growth resemble human NF1 symptoms. These and other dNf1 defects are relatively insensitive to manipulations that reduce Ras signaling strength but are suppressed by increasing signaling through the 3'-5' cyclic adenosine monophosphate (cAMP) dependent Protein Kinase A (PKA) pathway, or phenocopied by inhibiting this pathway. However, whether dNf1 affects cAMP/PKA signaling directly or indirectly remains controversial. To shed light on this issue we screened 486 1(st) and 2(nd) chromosome deficiencies that uncover >80% of annotated genes for dominant modifiers of the dNf1 pupal size defect, identifying responsible genes in crosses with mutant alleles or by tissue-specific RNA interference (RNAi) knockdown. Validating the screen, identified suppressors include the previously implicated dAlk tyrosine kinase, its activating ligand jelly belly (jeb), two other genes involved in Ras/ERK signal transduction and several involved in cAMP/PKA signaling. Novel modifiers that implicate synaptic defects in the dNf1 growth deficiency include the intersectin-related synaptic scaffold protein Dap160 and the cholecystokinin receptor-related CCKLR-17D1 drosulfakinin receptor. Providing mechanistic clues, we show that dAlk, jeb and CCKLR-17D1 are among mutants that also suppress a recently identified dNf1 neuromuscular junction (NMJ) overgrowth phenotype and that manipulations that increase cAMP/PKA signaling in adipokinetic hormone (AKH)-producing cells at the base of the neuroendocrine ring gland restore the dNf1 growth deficiency. Finally, supporting our previous contention that ALK might be a therapeutic target in NF1, we report that human ALK is expressed in cells that give rise to NF1 tumors and that NF1 regulated ALK/RAS/ERK signaling appears conserved in man.

BCG Artifact Removal for Reconstructing Full-scalp EEG inside the MR Scanner.

In simultaneous EEG/fMRI acquisition, the ballistocardiogram (BCG) artifact presents a major challenge for meaningful EEG signal interpretation and needs to be removed. This is very difficult, especially in continuous studies where BCG cannot be removed with averaging. In this study, we take advantage of a high-density EEG-cap and propose an integrated learning and inference approach to estimate the BCG contribution to the overall noisy recording. In particular, we present a special-designed experiment to enable a near-optimal subset selection scheme to identify a small set (20 out of 256 channels), and argue that in real-recording, BCG artifact signal from all channels can be estimated from this set. We call this new approach "Direct Recording Temporal Spatial Encoding" (DRTSE) to reflect these properties. In a preliminary evaluation, the DRTSE is combined with a direct subtraction and an optimization scheme to reconstruct the EEG signal. The performance was compared against the benchmark Optimal Basis Set (OBS) method. In the challenging nonevent-related EEG studies, the DRTSE method, with the optimization-based approach, yields an EEG reconstruction that reduces the normalized RMSE by approximately 13 folds, compared to OBS.

Automated diagnosis of epilepsy using EEG power spectrum.

Interictal electroencephalography (EEG) has clinically meaningful limitations in its sensitivity and specificity in the diagnosis of epilepsy because of its dependence on the occurrence of epileptiform discharges. We have developed a computer-aided diagnostic (CAD) tool that operates on the absolute spectral energy of the routine EEG and has both substantially higher sensitivity and negative predictive value than the identification of interictal epileptiform discharges. Our approach used a multilayer perceptron to classify 156 patients admitted for video-EEG monitoring. The patient population was diagnostically diverse; 87 were diagnosed with either generalized or focal seizures. The remainder of the patients were diagnosed with nonepileptic seizures. The sensitivity was 92% (95% confidence interval [CI] 85-97%) and the negative predictive value was 82% (95% CI 67-92%). We discuss how these findings suggest that this CAD can be used to supplement event-based analysis by trained epileptologists.

Forkhead box P1 overexpression and its clinicopathologic significance in peripheral T-cell lymphoma, not otherwise specified.

Forkhead box P1 protein is a transcription factor involved in cell signaling and regulation of gene expression and is essential for B-cell development. Forkhead box P1 overexpression has been associated with a worsened prognosis in some B-cell lymphomas. However, little is known about the clinicopathologic significance of forkhead box P1 in T-cell malignancies. In this study, immunohistochemistry for forkhead box P1 was performed in peripheral T-cell lymphoma, not otherwise specified, cases (n = 41), which were then divided into lower (n = 15) and higher (n = 26) forkhead box P1 expressers. Results of real-time quantitative reverse transcriptase polymerase chain reaction for forkhead box P1 messenger RNA supported the data on immunohistochemical forkhead box P1 expression. Forkhead box P1 overexpression in lymphoma cells was inversely associated with proliferation activity as evaluated by Ki-67 expression. Double immunostain for forkhead box P1 and a T-cell marker in normal lymph nodes showed forkhead box P1 signals in many of nonneoplastic T cells. Prognostic analysis showed that forkhead box P1 overexpression was associated with an improved overall survival of the patients with peripheral T-cell lymphoma, not otherwise specified, and was independent of the International Prognostic Index in multivariate analysis. Forkhead box P1 overexpression may be associated with less activated phenotype of the tumors and with a better prognosis in patients with peripheral T-cell lymphoma, not otherwise specified. The clinicopathologic significance of forkhead box P1 overexpression in peripheral T-cell lymphoma, not otherwise specified, may be different from that in B-cell lymphomas.

Characterization of chromosomal aberrations in thymic MALT lymphoma.

Mucosa-associated lymphoid tissue (MALT) lymphoma arising in the thymus is a rare disorder that shows a strong association with autoimmune disease. Several MALT-lymphoma-specific and -associated chromosomal abnormalities, including t(11;18), t(14;18), t(1;14), trisomy 3 and trisomy 18, are known to occur. The former translocation results in apoptosis inhibitor 2 gene (API2)-MALT lymphoma-associated translocation 1 (MALT1) fusion. In this study, we examined 14 cases of thymic MALT lymphomas for API2-MALT1 fusion using multiplex reverse transcription polymerase chain reaction and looked for trisomy 3, trisomy 18 and abnormalities of MALT1 and IGH genes using fluorescence in situ hybridization. Thymic MALT lymphoma cases had a high frequency of trisomy 3 (7/14 cases), a very low incidence of trisomy 18 (1/14) and no detectable MALT1-associated (0/13) or IGH-associated (0/13) gene abnormalities including t(11;18). A review of the literature showed that the pattern of chromosomal aberrations in thymic MALT lymphoma was similar to those of thyroid and salivary gland MALT lymphomas. Although frequently detected, trisomy 3 was not associated with any of the clinicopathological factors analyzed, suggesting that trisomy 3 may play a role in lymphoma development. In conclusion, the present study showed that thymic MALT lymphoma has a characteristic pattern of chromosomal aberrations that may be similar to those of other autoimmune-associated MALT lymphomas.

Parameter Selection in Mutual Information-Based Feature Selection in Automated Diagnosis of Multiple Epilepsies Using Scalp EEG.

Developing EEG-based computer aided diagnostic (CAD) tools would allow identification of epilepsy in individuals who have experienced possible seizures, yet such an algorithm requires efficient identification of meaningful features out of potentially more than 35,000 features of EEG activity. Mutual information can be used to identify a subset of minimally-redundant and maximally relevant (mRMR) features but requires a priori selection of two parameters: the number of features of interest and the number of quantization levels into which the continuous features are binned. Here we characterize the variance of cross-validation accuracy with respect to changes in these parameters for four classes of machine learning (ML) algorithms. This assesses the efficiency of combining mRMR with each of these algorithms by assessing when the variance of cross-validation accuracy is minimized and demonstrates how naive parameter selection may artificially depress accuracy. Our results can be used to improve the understanding of how feature selection interacts with four classes of ML algorithms and provide guidance for better a priori parameter selection in situations where an overwhelming number of redundant, noisy features are available for classification.

Analysis of API2-MALT1 fusion, trisomies, and immunoglobulin VH genes in pulmonary mucosa-associated lymphoid tissue lymphoma.

Pulmonary mucosa-associated lymphoid tissue lymphoma is unique in that chronic inflammation is rare and that API2-MALT1 fusion, resulting from t(11;18)(q21;q21), occurs frequently. In this study, we examined 20 cases for API2-MALT1 fusion using the multiplex reverse-transcription polymerase chain reaction and looked for trisomy 3, trisomy 18, and abnormalities of MALT1 and IGH genes using fluorescence in situ hybridization. In addition, we analyzed VH genes by subcloning of the monoclonal polymerase chain reaction products. Of 20 cases studied, we detected gene abnormalities in 16: API2-MALT1 fusion in 9, trisomy 3 in 5, trisomy 18 in 4, MALT1 abnormality in 13, and IGH abnormality in 1. MALT1 gene abnormalities were concordant with API2-MALT1 fusion or trisomy 18. One case showed API2-MALT1 fusion and trisomy 3. On detection of API2-MALT1 fusion and trisomies, we were able to divide our cases into 3 groups, API2-MALT1 positive, trisomy positive, and no detectable gene abnormality, suggesting that tumor development had processed along different genetic pathways. All 20 cases were analyzed for VH genes. Most of the VH genes selected by the lymphomas belonged to the VH3 family, but there was no restriction to any particular VH fragment. Of interest, VH genes were unmutated in 7 cases, suggesting that T-cell-independent extrafollicular B-cell maturation may be important in the development of this lymphoma. In addition, both mutated and unmutated tumor cases were found to carry the API2-MALT1 fusion and trisomy 3. This observation suggests that these gene abnormalities may occur in microenvironments found before or outside of follicular germinal centers.

Environmental enrichment reverses the impaired exploratory behavior and altered gene expression induced by early-life seizures.

Behavioral problems, school failure, and memory impairment are common among children with epilepsy. Currently, no effective treatment exists to promote recovery and neuron regeneration after seizures. To investigate the efficacy of environmental enrichment in reversing early-life seizure-induced changes in exploratory behavior and gene expression, we injected postnatal day 20 to 25 rats with kainic acid or saline and placed them either singly in a cage or as a group of eight in an enriched environment for 7 to 10 days. Exploratory behavior was quantified in an open field, and hippocampal gene analysis was performed on oligonucleotide microarrays. Exploratory behavior in kainic acid isolated rats were decreased in open field, whereas kainic acid rats exposed to an enriched environment behaved similarly to controls (n = 37, analysis of variance, P < .001). Correlated with an improvement in behavior, genes involved in synaptic plasticity and memory consolidation, such as Arc, Homer1a, and Egr1, were significantly increased in rats exposed to environmental enrichment. Real-time quantitative reverse transcriptase-polymerase chain reaction confirmed our microarray data on select genes. Our results provide an experimental basis for promoting enriching education programs for children with epilepsy.