Ensemble Learning Method for In-Hospital Cardiac Arrest Prediction.

Cardiac arrest prediction for multivariate time series data have been developed and obtained high precision performance. However, these algorithms still did not achieved high sensitivity and suffer from a high false-alarm. Therefore, we propose a ensemble approach for prediction satisfying precision-recall result compared than other machine learning methods. As a result, our proposed method obtained an overall area under precision-recall curve of 46.7%. It is possible to more accurately respond rapidly cardiac arrest event.

Explainable Artificial Intelligence Warning Model Using an Ensemble Approach for In-Hospital Cardiac Arrest Prediction: Retrospective Cohort Study.

BACKGROUND: Cardiac arrest (CA) is the leading cause of death in critically ill patients. Clinical research has shown that early identification of CA reduces mortality. Algorithms capable of predicting CA with high sensitivity have been developed using multivariate time series data. However, these algorithms suffer from a high rate of false alarms, and their results are not clinically interpretable. OBJECTIVE: We propose an ensemble approach using multiresolution statistical features and cosine similarity-based features for the timely prediction of CA. Furthermore, this approach provides clinically interpretable results that can be adopted by clinicians. METHODS: Patients were retrospectively analyzed using data from the Medical Information Mart for Intensive Care-IV database and the eICU Collaborative Research Database. Based on the multivariate vital signs of a 24-hour time window for adults diagnosed with heart failure, we extracted multiresolution statistical and cosine similarity-based features. These features were used to construct and develop gradient boosting decision trees. Therefore, we adopted cost-sensitive learning as a solution. Then, 10-fold cross-validation was performed to check the consistency of the model performance, and the Shapley additive explanation algorithm was used to capture the overall interpretability of the proposed model. Next, external validation using the eICU Collaborative Research Database was performed to check the generalization ability. RESULTS: The proposed method yielded an overall area under the receiver operating characteristic curve (AUROC) of 0.86 and area under the precision-recall curve (AUPRC) of 0.58. In terms of the timely prediction of CA, the proposed model achieved an AUROC above 0.80 for predicting CA events up to 6 hours in advance. The proposed method simultaneously improved precision and sensitivity to increase the AUPRC, which reduced the number of false alarms while maintaining high sensitivity. This result indicates that the predictive performance of the proposed model is superior to the performances of the models reported in previous studies. Next, we demonstrated the effect of feature importance on the clinical interpretability of the proposed method and inferred the effect between the non-CA and CA groups. Finally, external validation was performed using the eICU Collaborative Research Database, and an AUROC of 0.74 and AUPRC of 0.44 were obtained in a general intensive care unit population. CONCLUSIONS: The proposed framework can provide clinicians with more accurate CA prediction results and reduce false alarm rates through internal and external validation. In addition, clinically interpretable prediction results can facilitate clinician understanding. Furthermore, the similarity of vital sign changes can provide insights into temporal pattern changes in CA prediction in patients with heart failure-related diagnoses. Therefore, our system is sufficiently feasible for routine clinical use. In addition, regarding the proposed CA prediction system, a clinically mature application has been developed and verified in the future digital health field.

Machine Learning Algorithms Predict Successful Weaning From Mechanical Ventilation Before Intubation: Retrospective Analysis From the Medical Information Mart for Intensive Care IV Database.

BACKGROUND: The prediction of successful weaning from mechanical ventilation (MV) in advance of intubation can facilitate discussions regarding end-of-life care before unnecessary intubation. OBJECTIVE: We aimed to develop a machine learning-based model that predicts successful weaning from ventilator support based on routine clinical and laboratory data taken before or immediately after intubation. METHODS: We used the Medical Information Mart for Intensive Care IV database, which is an open-access database covering 524,740 admissions of 382,278 patients in Beth Israel Deaconess Medical Center, United States, from 2008 to 2019. We selected adult patients who underwent MV in the intensive care unit (ICU). Clinical and laboratory variables that are considered relevant to the prognosis of the patient in the ICU were selected. Data collected before or within 24 hours of intubation were used to develop machine learning models that predict the probability of successful weaning within 14 days of ventilator support. Developed models were integrated into an ensemble model. Performance metrics were calculated by 5-fold cross-validation for each model, and a permutation feature importance and Shapley additive explanations analysis was conducted to better understand the impacts of individual variables on outcome prediction. RESULTS: Of the 23,242 patients, 19,025 (81.9%) patients were successfully weaned from MV within 14 days. Using the preselected 46 clinical and laboratory variables, the area under the receiver operating characteristic curve of CatBoost classifier, random forest classifier, and regularized logistic regression classifier models were 0.860 (95% CI 0.852-0.868), 0.855 (95% CI 0.848-0.863), and 0.823 (95% CI 0.813-0.832), respectively. Using the ensemble voting classifier using the 3 models above, the final model revealed the area under the receiver operating characteristic curve of 0.861 (95% CI 0.853-0.869), which was significantly better than that of Simplified Acute Physiology Score II (0.749, 95% CI 0.742-0.756) and Sequential Organ Failure Assessment (0.588, 95% CI 0.566-0.609). The top features included lactate and anion gap. The model's performance achieved a plateau with approximately the top 21 variables. CONCLUSIONS: We developed machine learning algorithms that can predict successful weaning from MV in advance to intubation in the ICU. Our models can aid the appropriate management for patients who hesitate to decide on ventilator support or meaningless end-of-life care.

In vivo glucose metabolism and glutamate levels in mGluR5 knockout mice: a multimodal neuroimaging study using [18F]FDG microPET and MRS.

BACKGROUND: Perturbed functional coupling between the metabotropic glutamate receptor-5 (mGluR5) and N-methyl-D-aspartate (NMDA) receptor-mediated excitatory glutamatergic neurotransmission may contribute to the pathophysiology of psychiatric disorders such as schizophrenia. We aimed to establish the functional interaction between mGluR5 and NMDA receptors in brain of mice with genetic ablation of the mGluR5. METHODS: We first measured the brain glutamate levels with magnetic resonance spectroscopy (MRS) in mGluR5 knockout (KO) and wild-type (WT) mice. Then, we assessed brain glucose metabolism with [18F]fluorodeoxyglucose ([18F]FDG) positron emission tomography before and after the acute administration of an NMDA antagonist, MK-801 (0.5 mg/kg), in the same mGluR5 KO and WT mice. RESULTS: Between-group comparisons showed no significant differences in [18F]FDG standardized uptake values (SUVs) in brain of mGluR5 KO and WT mice at baseline, but widespread reductions in mGluR5 KO mice compared to WT mice after MK-801 administration (p < 0.05). The baseline glutamate levels did not differ significantly between the two groups. However, there were significant negative correlations between baseline prefrontal glutamate levels and regional [18F]FDG SUVs in mGluR5 KO mice (p < 0.05), but no such correlations in WT mice. Fisher's Z-transformation analysis revealed significant between-group differences in these correlations (p < 0.05). CONCLUSIONS: This is the first multimodal neuroimaging study in mGluR5 KO mice and the first report on the association between cerebral glucose metabolism and glutamate levels in living rodents. The results indicate that mGluR5 KO mice respond to NMDA antagonism with reduced cerebral glucose metabolism, suggesting that mGluR5 transmission normally moderates the net effects of NMDA receptor antagonism on neuronal activity. The negative correlation between glutamate levels and glucose metabolism in mGluR5 KO mice at baseline may suggest an unmasking of an inhibitory component of the glutamatergic regulation of neuronal energy metabolism.

In vivo metabotropic glutamate receptor 5 availability-associated functional connectivity alterations in drug-naïve young adults with major depression.

There has been increasing interest in glutamatergic neurotransmission as a putative underlying mechanism of depressive disorders. We performed [11C]ABP688 positron emission tomography (PET) and resting-state functional magnetic resonance imaging (rs-fMRI) in drug-naïve young adult patients with major depression to examine alterations in metabotropic glutamate receptor-5 (mGluR5) availability, and to investigate their functional significance relating to neural systems-level changes in major depression. Sixteen psychotropic drug-naïve patients with major depression without comorbidity (median age: 22.8 years) and fifteen matched healthy controls underwent [11C]ABP688 PET imaging and 3-T MRI. For mGluR5 availability, we quantified [11C]ABP688 binding potential (BPND) using the simplified reference tissue model. Seed-based functional connectivity analysis was performed using rs-fMRI data with regions derived from quantitative [11C]ABP688 PET analysis as seeds. In region-of-interest (ROI)-based and voxel-based analyses, the [11C]ABP688 BPND was significantly lower in patients than in controls in the prefrontal cortex ROI and in voxel clusters within the prefrontal, temporal, and parietal cortices, and supramarginal gyrus. The [11C]ABP688 BPND seed-based functional connectivity analysis showed significantly less negative connectivity from the inferior parietal cortex seed to the fusiform gyrus and inferior occipital cortex in patients than in controls. The correlation patterns between [11C]ABP688 BPND and functional connectivity strength (ß) for the superior prefrontal cortex seed were opposite in the depression and control groups. In conclusion, using a novel approach combining [11C]ABP688 PET and rs-fMRI analyses, our study provides a first evidence of lower mGluR5 availability and related functional connectivity alterations in drug-naïve young adults with major depression without comorbidity.

Changes in network connectivity during motor imagery and execution.

BACKGROUND: Recent studies of functional or effective connectivity in the brain have reported that motor-related brain regions were activated during motor execution and motor imagery, but the relationship between motor and cognitive areas has not yet been completely understood. The objectives of our study were to analyze the effective connectivity between motor and cognitive networks in order to define network dynamics during motor execution and motor imagery in healthy individuals. Second, we analyzed the differences in effective connectivity between correct and incorrect responses during motor execution and imagery using dynamic causal modeling (DCM) of electroencephalography (EEG) data. METHOD: Twenty healthy subjects performed a sequence of finger tapping trials using either motor execution or motor imagery, and the performances were recorded. Changes in effective connectivity between the primary motor cortex (M1), supplementary motor area (SMA), premotor cortex (PMC), and dorsolateral prefrontal cortex (DLPFC) were estimated using dynamic causal modeling. Bayesian model averaging with family-level inference and fixed-effects analysis was applied to determine the most likely connectivity model for these regions. RESULTS: Motor execution and imagery showed inputs to distinct brain regions, the premotor cortex and the supplementary motor area, respectively. During motor execution, the coupling strength of a feedforward network from the DLPFC to the PMC was greater than that during motor imagery. During motor imagery, the coupling strengths of a feedforward network from the PMC to the SMA and of a feedback network from M1 to the PMC were higher than that during motor execution. In imagined movement, although there were connectivity differences between correct and incorrect task responses, each motor imagery task that included correct and incorrect responses showed similar network connectivity characteristics. Correct motor imagery responses showed connectivity from the PMC to the DLPFC, while the incorrect responses had characteristic connectivity from the SMA to the DLPFC. CONCLUSIONS: These findings provide an understanding of effective connectivity between motor and cognitive areas during motor execution and imagery as well as the basis for future connectivity studies for patients with stroke.

Effect of Dual-Mode and Dual-Site Noninvasive Brain Stimulation on Freezing of Gait in Patients With Parkinson Disease.

OBJECTIVE: To investigate the effect of dual-mode noninvasive brain stimulation (NIBS) with high-frequency repetitive transcranial magnetic stimulation (rTMS) over the primary motor cortex of the lower leg and anodal transcranial direct current stimulation (tDCS) over the left dorsolateral prefrontal cortex compared with rTMS alone in patients with Parkinson disease (PD) with freezing of gait (FOG). DESIGN: Randomized, double-blind, controlled study. SETTING: Outpatient rehabilitation clinics. PARTICIPANTS: Patients diagnosed as having PD with FOG (N=32). INTERVENTIONS: Patients in the dual-mode group underwent 5 consecutive daily sessions of dual-mode NIBS with high-frequency rTMS and tDCS simultaneously, whereas patients in the rTMS group underwent high-frequency rTMS and sham tDCS. MAIN OUTCOME MEASURES: Assessments of FOG and motor, ambulatory, and cognitive function were performed 3 times: at baseline before NIBS, immediately after NIBS, and 1 week after cessation of NIBS. RESULTS: Serious adverse effects were not observed in either group. Significant changes over time were observed in FOG, motor function, and ambulatory function in each group; however, there was no significant difference between the 2 groups. Executive function showed significant improvement after NIBS only in the dual-mode group. CONCLUSIONS: These results suggest the potential for dual-mode NIBS to modulate 2 different cortices simultaneously. Dual-mode NIBS might be considered a novel therapeutic approach for patients with PD.

Effects of Bilateral Repetitive Transcranial Magnetic Stimulation on Post-Stroke Dysphagia.

BACKGROUND: Optimal protocol of repetitive transcranial magnetic stimulation (rTMS) on post-stroke dysphagia remains uncertain with regard to its clinical efficacy. OBJECTIVE: The aim of the present study is to investigate the effects of high-frequency rTMS at the bilateral motor cortices over the cortical representation of the mylohyoid muscles in the patients with post-stroke dysphagia. METHODS: This study was a single-blind, randomized controlled study with a blinded observer. Thirty-five stroke patients were randomly divided into three intervention groups: the bilateral stimulation group, the unilateral stimulation group, and the sham stimulation group. For the bilateral stimulation group, 500 pulses of 10 Hz rTMS over the ipsilesional and 500 pulses of 10 Hz rTMS over the contralesional motor cortices over the cortical areas that project to the mylohyoid muscles were administered daily for 2 consecutive weeks. For the unilateral stimulation group, 500 pulses of 10 Hz rTMS over the ipsilesional motor cortex over the cortical representation of the mylohyoid muscle and the same amount of sham rTMS over the contralesional hemisphere were applied. For the sham stimulation group, sham rTMS was applied at the bilateral motor cortices. Clinical swallowing function and videofluoroscopic swallowing studies were assessed before the intervention (T0), immediately after the intervention (T1) and 3 weeks after the intervention (T2) using Clinical Dysphagia Scale (CDS), Dysphagia Outcome and Severity Scale (DOSS), Penetration Aspiration Scale (PAS), and Videofluoroscopic Dysphagia Scale (VDS). RESULTS: There were significant time and intervention interaction effects in the CDS, DOSS, PAS, and VDS scores (p < 0.05). In the direct comparison of the changes in the swallowing parameters among the three groups, the change in CDS scores at T1 and T2 showed a significantly higher improvement in the bilateral simulation group than in two other groups (p < 0.05). There was a significantly larger change in the DOSS, PAS, and VDS scores at T1 in the bilateral stimulation group than in two other groups (p < 0.05). CONCLUSIONS: The results of the present study provide substantial evidence that 10 Hz rTMS at the bilateral motor cortices over the cortical areas projecting to the mylohyoid muscles is effective as an additional treatment strategy to traditional dysphagia therapies.

Effect of cumulative repetitive transcranial magnetic stimulation on freezing of gait in patients with atypical Parkinsonism: A pilot study.

OBJECTIVE: To investigate the potential of cumulative high-frequency repetitive transcranial magnetic stimulation (HF-rTMS) on freezing of gait in atypical Parkinsonism. DESIGN: Randomized, single-blinded, crossover study with a blinded observer. PARTICIPANTS: Eight patients with atypical Parkinsonism. METHODS: All participants received HF-rTMS over the lower leg primary motor cortex (M1-LL) for 5 consecutive days. Alternative sham stimulation was also administered with a 2-week wash-out period. Freezing of Gait Questionnaire (FOG-Q), turn steps in the modified Standing Start 180° Turn Test, the Timed Up and Go (TUG) task, and the Unified Parkinson's Disease Rating Scale part III (UPDRS-III) were performed before, after, and one week after rTMS. RESULTS: All participants completed this study without any significant adverse effects. FOG-Q and turn steps revealed significant improvements over time in the rTMS compared with the sham stimulation (χ2=6.067, p=0.048 and χ2=9.083, p=0.011). In addition, the TUG task and UPDRS-III showed significant improvements over time in the rTMS compared with the sham stimulation (χ2=7.200, p=0.02 and χ2=7.000, p=0.030). CONCLUSION: Cumulative HF-rTMS over the M1-LL might be effective for improving freezing of gait in patients with atypical Parkinsonism. Further investigation with a large number of participants is needed to clarify the effects of HF- rTMS on freezing of gait in atypical Parkinsonism.

Usefulness of Transcranial Magnetic Stimulation to Assess Motor Function in Patients With Parkinsonism.

OBJECTIVE: To investigate the clinical significance of upper and lower extremity transcranial magnetic stimulation (TMS)-induced motor evoked potentials (MEPs) in patients with parkinsonism. METHODS: Twenty patients (14 men, 6 women; mean age 70.5±9.1 years) suffering from parkinsonism were included in this study. All participants underwent single-pulse TMS session to assess the corticospinal excitability of the upper and lower extremity motor cortex. The resting motor threshold (RMT) was defined as the lowest stimulus intensity able to evoke MEPs of an at least 50 µV peak-to-peak amplitude in 5 of 10 consecutive trials. Five sweeps of MEPs at 120% of the RMT were performed, and the mean amplitude and latency of the MEPs were calculated. Patients were also assessed using the Unified Parkinson's Disease Rating Scale part III (UPDRS-III) and the 5-meter Timed Up and Go (5m-TUG) test. RESULTS: There was a significant positive correlation between the RMTs of MEPs in the upper and lower extremities (r=0.612, p=0.004) and between the amplitude of MEPs in the upper and lower extremities (r=0.579, p=0.007). The RMT of upper extremity MEPs showed a significant negative relationship with the UPDRS-III score (r=-0.516, p=0.020). In addition, RMTs of lower extremity MEPs exhibited a negative relationship with the UPDRS-III score, but the association was not statistically significant (r=-406, p=0.075). CONCLUSION: These results indicated that the RMT of MEPs reflect the severity of motor dysfunction in patients with parkinsonism. MEP is a potential quantitative, electrodiagnostic method to assess motor function in patients with parkinsonism.

Efficacy of cumulative high-frequency rTMS on freezing of gait in Parkinson's disease.

PURPOSE: Freezing of gait (FOG) affects mobility and balance seriously. Few reports have investigated the effects of repetitive transcranial magnetic stimulation (rTMS) on FOG in Parkinson's disease (PD). We investigated the efficacy of high-frequency rTMS for the treatment of FOG in PD. METHODS: Seventeen patients diagnosed with PD were recruited in a randomized, double-blinded, cross-over study. We applied high frequency rTMS (90% of resting motor threshold, 10 Hz, 1,000 pulses) over the lower leg primary motor cortex of the dominant hemisphere (M1-LL) for five sessions in a week. We also administered alternative sham stimulation with a two-week wash out period. The primary outcomes were measured before, immediately after, and one week after the intervention using the Standing Start 180° Turn Test (SS-180) with video analysis and the Freezing of Gait Questionnaire (FOG-Q). The secondary outcome measurements consisted of Timed Up and Go (TUG) tasks and the Unified Parkinson's Disease Rating Scale part III (UPDRS-III). Motor cortical excitability was also evaluated. RESULTS: There were significant improvements in the step required to complete the SS-180 and FOG-Q in the rTMS condition compared to the sham condition, and the effects continued for a week. The TUG and UPDRS-III also showed significant ameliorations over time in the rTMS condition. The MEP amplitude at 120% resting motor threshold and intracortical facilitation also increased after real rTMS condition. CONCLUSIONS: High frequency rTMS over the M1-LL may serve as an add-on therapy for improving FOG in PD.