Explainable Machine Learning for Coma Outcome Prediction Based on Structural and Functional Brain MRI.

BackgroundThe urinary albumin/creatinine ratio (ACR) is a significant neurologic prognostic predictor in patients with aneurysmal subarachnoid hemorrhage (SAH). B-type natriuretic peptide (BNP) plays an important role in body fluid regulation in patients with SAH. The present study was performed to determine whether ACR was independent predictor for unfavorable neurological outcome and ACR was associated with increased N-terminal pro-BNP (NT-pro-BNP) after SAH.MethodsWe studied 61 patients undergoing surgery who were admitted within 48 h after aneurysmal SAH onset between July 2008 and June 2010. Hunt and Hess grade and Fisher grade were recorded at admission. The Glasgow Coma Scale (GCS) score was calculated at admission and daily for seven postoperative days. Arterial blood was sampled at admission and for seven postoperative days to determine the PaO2/FIO2 ratio, C-reactive protein level, troponin I level, and NT-pro-BNP level. Urine was sampled at admission and daily for seven postoperative days to determine ACR and vanillylmandelic acid/creatinine ratio (VMACR). Neurological outcomes were assessed at hospital discharge by using the Glasgow Outcome Scale. Receiver operating characteristic curves were constructed for the predictive variables of unfavorable neurological outcomes, and the area under the curve (AUC) was determined. Multivariate logistic regression analyses were performed for the significant predictors of unfavorable neurological outcomes after SAH. Associations with NT-pro-BNP were evaluated by using the Spearman rank correlation test.ResultsOf the 61 patients, 24 had unfavorable outcomes. The prevalence rate of microalbuminuria was 85 % (52/61). The highest NT-pro-BNP levels were above the normal range in 57 of 61 patients (93 %).According to the AUC, the Hunt and Hess grade, GCS score, the highest ACR, and highest VMACR were significant predictors of neurological outcome. Multivariate logistic regression analyses showed that the highest ACR and Hunt and Hess grade are independent prognostic predictors of unfavorable neurological outcomes. The highest NT-pro-BNP significantly correlated with the highest troponin I, highest ACR, and VMACR on admission.ConclusionsThe highest ACR is an independent prognostic predictor of unfavorable neurological outcomes after SAH. Moreover, plasma NT-pro-BNP elevation may be associated with the development of microalbuminuria.

<bold>Objective</bold> To explore the effect of repetitive transcranial direct current combined with rehabilitation therapy on promoting consciousness and improving spasticity in patients with consciousness disorder. <bold>Methods</bold> A retrospective analysis was performed on 42 patients with disturbance of consciousness，whom hospitalized in rehabilitation medicine department of Lanzhou University Second Hospital from April 2017 to April 2021. They were divided into observation group and control group according to whether they received transcranial direct current stimulation （tDCS），21 cases in each group. The control group were treated with conventional rehabilitation treatment and the observation group were treated with conventional rehabilitation therapy and tDCS，the anode was placed in the left primary motor cortex （M1） and the cathode was placed in the right shoulder region，current intensity of 2 mA，once daily，20 min once a day，5 times per week，for 6 weeks. Before and after treatment，the level of consciousness was assessed by the Coma Recovery Scale-Revised （CRS-R） and Glasgow Coma Scale （GCS），the Modified Ashworth Scale （MAS） was used to assess changes muscle tone of the right flexor elbow muscle. <bold>Results</bold> Before treatment，there were no significant difference in CRS-R and GCS scores between the two groups （<italic>P</italic>＞0.05）；After treatment，the scores of the two groups were improved，and the CRS-R and GCS scores of the observation group were better than the control group，the differences were statistically significant （<italic>P</italic>&amp;lt;0.05）. In terms of clinical efficacy，there were 13 （38.10%） responders and 8 （61.90%） non responders in the observation group. In the control group，there were 2 （9.50%） patients with responded and 19 （90.50%） patients who did not respond，the difference between the two groups was significant difference （<italic>P</italic>&amp;lt;0.05）. As for MAS score，there was no significant difference in MAS score of right flexor elbow muscle between the two groups before treatment （<italic>P</italic>＞0.05）; There was no significant difference in MAS score of bilateral flexor elbow muscle in the control group before and after treatment （<italic>P</italic>＞0.05），but there was significant difference in the observation group before and after treatment （<italic>P</italic>&amp;lt;0.05）. Comparison among groups，MAS score of flexor elbow muscle in the observation group was better than the control group，the difference was significant difference （<italic>P</italic>&amp;lt;0.05）. <bold>Conclusion</bold> Repetitive transcranial direct current stimulation combined with rehabilitation therapy can effectively improve the level of consciousness and spasticity of patients with consciousness disorder.

Several models exist to predict trauma center need in the prehospital setting; however, there is lack of simple clinical tools to predict the need for ICU admission and mortality in trauma patients. The aim of our study was to develop a simple clinical tool that can be used with ease in the prehospital or emergency setting and can reliably predict the need for ICU admission and mortality in trauma patients. We abstracted one year of National Trauma Data Bank for all patients aged ≥ 18 years. Transferred patients and those dead on arrival were excluded. Patient demographics, injury parameters, vital signs, and Glasgow Coma Scale (GCS) were recorded. Our primary outcome measures were mortality and ICU admission. Logistic regression analysis was performed using three variables (age > 55 years, shock index (SI) > 1, and GCS score) to determine the appropriate weights for predicting mortality. Appropriate weights derived from regression analysis were used to construct a simple SI, age, and GCS (SAG) score, and associated mortality and ICU admissions were calculated for three different risk groups (low, intermediate, and high). A total of 281,522 patients were included. The mean age was 47 ± 20 years, and 65 per cent were male. The overall mortality rate was 2.9 per cent, and the rate of ICU admission was 28.7 per cent. The SAG score was constructed using weights derived from regression analysis for age ≤ 55 years (4 points), SI < 1 (3 points), and GCS (3-15 points). The median [IQR] SAG score was 21 [18-22]. The area under the receiver operating curve [95% Confidence Interval (CI)] of the SAG score for predicting mortality and ICU admission was 0.873 [0.870-0.877] and 0.644 [0.642-0.647], respectively. Each 1-point increase in the SAG score was associated with 18 per cent lower odds of mortality (odds ratio [95% CI]: 0.822 [0.820-0.825]) and 10 per cent lower odds of ICU admission (odds ratio [95% CI]: 0.901 [0.899-0.902]). The SAG score is a simple clinical tool derived from variables that can be assessed with ease during the initial evaluation of trauma patients. It provides a rapid assessment and can reliably predict mortality and need for ICU admission in trauma patients. This simple tool may allow early resource mobilization possibly even before the arrival of the patient.

ObjectiveTo investigate the short-term effect of music therapy combined with binaural frequency difference therapy on patients with consciousness disorder.Materials and methodsNinety patients with definite diagnosis of disorders of consciousness (DOC) were selected. These patients were randomly divided into control group, experiment 1 group and experiment 2 group, with 30 patients in each group. The control group was treated with routine clinical treatment and rehabilitation. In experiment 1 group, music therapy was added to the control group. In experimental group 2, music therapy combined with binaural α frequency difference therapy was added to the control group. All patients were assessed before and after 30 treatments. The assessment items included Glasgow Coma Scale (GCS), Coma Recovery Scale revised (CRS-R), electroencephalogram (EEG), upper somatosensory evoked potential (USEP), and brainstem auditory evoked potential (BAEP).ResultsBefore treatment, there were no significant differences in GCS score, CRS-R score, USEP, BAEP, and EEG scores among the three groups (P > 0.05). After 30 times of treatment, GCS score, CRS-R score, USEP, BAEP, and EEG scores in 3 groups were significantly higher than those before treatment (P < 0.05), and experimental group 2 >experimental group 1 >control group (P < 0.05). And the consciousness rate of experimental group 2 was better than experimental group 1, experimental group 1 was better than the control group and the difference was statistically significant (P < 0.05).ConclusionMusic therapy combined with binaural α frequency difference therapy is more effective in stimulating DOC patients.

Purpose: Maternal mortality remained to persist in many developing countries. India being the most populous developing country has several cultural differences and beliefs on health systems and may have disparities in receiving proper maternal health care. High data availability and under-utilization of data centric decision making is a key reason for ineffective performance of health systems. Enacting machine learning on such data to aid in sub-divisional policy formulation to address area level problems will eradicate major disparities in recipients of health services. Methods: Glass box machine learning models are trained on the data to obtain importance of features in defining the maternal mortality of a district. Furthermore, black box machine learning models are trained with hyper-parameter tuning and best model is chosen to perform explainable machine learning to generate explanations for each district prediction. A hybrid explainable machine learning approach is proposed on black-box machine learning models where Shapley Additive Explanation and Local Interpretable Model-agnostic Explanations are combined to generate final explanations. Results: There may be several differences even among nearby districts. Health Management Information System data is analyzed with help of Machine Learning techniques and Explainable Machine Learning techniques are used on the trained models to evaluate the contributing factors for each district. Conclusion: The factors that are specific to each district can help in formulating region specific health policies that minimize the disparities of progress of preventing maternal mortality over the districts of India. The paper has highlighted the advantages of using explainable machine learning in extracting complicated patterns of the data.

The Coma Recovery Scale-Revised (CRS-R) is the most recommended clinical tool to examine the neurobehavioral condition of individuals with disorders of consciousness (DOCs). Different studies have investigated the prognostic value of the information provided by the conventional administration of the scale, while other measures derived from the scale have been proposed to improve the prognosis of DOCs. However, the heterogeneity of the data used in the different studies prevents a reliable comparison of the identified predictors and measures. This study investigates which information derived from the CRS-R provides the most reliable prediction of both the clinical diagnosis and recovery of consciousness at the discharge of a long-term neurorehabilitation program. Retrospective observational multisite study. The enrollment was performed in three neurorehabilitation facilities of the same hospital network. A total of 171 individuals with DOCs admitted to an inpatient neurorehabilitation program for a minimum of 3 months were enrolled. Machine learning classifiers were trained to predict the clinical diagnosis and recovery of consciousness at discharge using clinical confounders and different metrics extracted from the CRS-R scale. Results showed that the neurobehavioral state at discharge was predicted with acceptable and comparable predictive value with all the indices and measures derived from the CRS-R, but for the clinical diagnosis and the Consciousness Domain Index, and the recovery of consciousness was predicted with higher accuracy and similarly by all the investigated measures, with the exception of initial clinical diagnosis. Interestingly, the total score in the CRS-R and, especially, the total score in its subscales provided the best overall results, in contrast to the clinical diagnosis, which could indicate that a comprehensive measure of the clinical diagnosis rather than the condition of the individuals could provide a more reliable prediction of the neurobehavioral progress of individuals with prolonged DOC. The results of this work have important implications in clinical practice, offering a more accurate prognosis of patients and thus giving the possibility to personalize and optimize the rehabilitation plan of patients with DoC using low-cost and easily collectable information.

BackgroundProlonged Disorders of Consciousness (PDOC) resulting from severe acquired brain injury can lead to complex disabilities that make diagnosis challenging. The role of machine learning (ML) in diagnosing PDOC states and identifying intervention strategies is relatively under-explored, having focused on predicting mortality and poor outcome. This study aims to: (a) apply ML techniques to predict PDOC diagnostic states from variables obtained from two non-invasive neurobehavior assessment tools; and (b) apply network analysis for guiding possible intervention strategies.MethodsThe Coma Recovery Scale-Revised (CRS-R) is a well-established tool for assessing patients with PDOC. More recently, music has been found to be a useful medium for assessment of coma patients, leading to the standardization of a music-based assessment of awareness: Music Therapy Assessment Tool for Awareness in Disorders of Consciousness (MATADOC). CRS-R and MATADOC data were collected from 74 PDOC patients aged 16–70 years at three specialist centers in the USA, UK and Ireland. The data were analyzed by three ML techniques (neural networks, decision trees and cluster analysis) as well as modelled through system-level network analysis.ResultsPDOC diagnostic state can be predicted to a relatively high level of accuracy that sets a benchmark for future ML analysis using neurobehavioral data only. The outcomes of this study may also have implications for understanding the role of music therapy in interdisciplinary rehabilitation to help patients move from one coma state to another.ConclusionsThis study has shown how ML can derive rules for diagnosis of PDOC with data from two neurobehavioral tools without the need to harvest large clinical and imaging datasets. Network analysis using the measures obtained from these two non-invasive tools provides novel, system-level ways of interpreting possible transitions between PDOC states, leading to possible use in novel, next-generation decision-support systems for PDOC.

Despite advances in resting state functional magnetic resonance imaging investigations, clinicians remain with the challenge of how to implement this paradigm on an individualized basis. Here, we assessed the clinical relevance of resting state functional magnetic resonance imaging acquisitions in patients with disorders of consciousness by means of a systems-level approach. Three clinical centres collected data from 73 patients in minimally conscious state, vegetative state/unresponsive wakefulness syndrome and coma. The main analysis was performed on the data set coming from one centre (Liège) including 51 patients (26 minimally conscious state, 19 vegetative state/unresponsive wakefulness syndrome, six coma; 15 females; mean age 49 ± 18 years, range 11-87; 16 traumatic, 32 non-traumatic of which 13 anoxic, three mixed; 35 patients assessed >1 month post-insult) for whom the clinical diagnosis with the Coma Recovery Scale-Revised was congruent with positron emission tomography scanning. Group-level functional connectivity was investigated for the default mode, frontoparietal, salience, auditory, sensorimotor and visual networks using a multiple-seed correlation approach. Between-group inferential statistics and machine learning were used to identify each network's capacity to discriminate between patients in minimally conscious state and vegetative state/unresponsive wakefulness syndrome. Data collected from 22 patients scanned in two other centres (Salzburg: 10 minimally conscious state, five vegetative state/unresponsive wakefulness syndrome; New York: five minimally conscious state, one vegetative state/unresponsive wakefulness syndrome, one emerged from minimally conscious state) were used to validate the classification with the selected features. Coma Recovery Scale-Revised total scores correlated with key regions of each network reflecting their involvement in consciousness-related processes. All networks had a high discriminative capacity (>80%) for separating patients in a minimally conscious state and vegetative state/unresponsive wakefulness syndrome. Among them, the auditory network was ranked the most highly. The regions of the auditory network which were more functionally connected in patients in minimally conscious state compared to vegetative state/unresponsive wakefulness syndrome encompassed bilateral auditory and visual cortices. Connectivity values in these three regions discriminated congruently 20 of 22 independently assessed patients. Our findings point to the significance of preserved abilities for multisensory integration and top-down processing in minimal consciousness seemingly supported by auditory-visual crossmodal connectivity, and promote the clinical utility of the resting paradigm for single-patient diagnostics.

Objective.Functional network connectivity (FNC) estimated from resting-state functional magnetic resonance imaging showed great information about the neural mechanism in different brain disorders. But previous research has mainly focused on standard statistical learning approaches to find FNC features separating patients from control. While machine learning models can improve classification accuracy, they often lack interpretability, making it difficult to understand how they arrive at their decisions.Approach.Explainable machine learning helps address this issue by identifying which features contribute most to the model’s predictions. In this study, we introduce a novel framework leveraging SHapley Additive exPlanations (SHAPs) to identify crucial FNC features distinguishing between two distinct population classes.Main results.Initially, we validate our approach using synthetic data. Subsequently, applying our framework, we ascertain FNC biomarkers distinguishing between, controls and schizophrenia (SZ) patients with accuracy of 81.04% as well as middle aged adults and old aged adults with accuracy 71.38%, respectively, employing random forest, XGBoost, and CATBoost models.Significance.Our analysis underscores the pivotal role of the cognitive control network (CCN), subcortical network (SCN), and somatomotor network in discerning individuals with SZ from controls. In addition, our platform found CCN and SCN as the most important networks separating young adults from older.

Management of phosphate mine waste rock piles (PWRPs) is a critical challenge in the mining industry, particularly in regions like Morocco, which holds the world&amp;#8217;s largest phosphate reserves. To this end, there is a need for an approach that focuses on real-time monitoring of waste rock heterogeneity, enabling more efficient resource recovery and environmental management. This study proposes a novel, multi-scale approach that integrates hyperspectral imaging, field spectroscopy, and explainable machine learning (XML) to characterize and map the mineralogical diversity of PWRPs at the Benguerir mine. &amp;#160;A total of 103 samples were collected from waste rock piles across an area of approximately 60 km&amp;#178;, representing the full spectrum of mineralogical variability. Handheld X-ray fluorescence (XRF) analysis was conducted on the all the samples and revealed the dominance of SiO&amp;#8322; (29.51 wt% &amp;#177; 12.42), CaO (30.16 wt% &amp;#177; 10.17), and P&amp;#8322;O&amp;#8325; (7.23 wt% &amp;#177; 4.21). These XRF analyses indicated the presence of silicate, carbonate, and phosphate-bearing materials. These findings were complemented by both PRISMA hyperspectral imaging, which captured spectral data across the visible to shortwave infrared (VSWIR) range. precise calibration and validation of the remote sensing outputs were conducted using field spectroscopy using the ASD FieldSpec 4 spectroradiometer.To address the complexity of the spectral data, we developed an explainable machine learning framework based on SHapley Additive exPlanations (SHAP) and Convolutional Neural Networks (CNN). This framework not only improved classification accuracy (achieving 0.92 overall accuracy) but also provided interpretable insights into the spectral features driving mineral identification. Our results showed that the used model successfully differentiated four main waste rock categories: carbonate-rich, phosphate-rich, clay-dominated, and siliceous materials. The resulting maps offer a practical tool for real-time waste management and resource recovery. For instance, carbonate-rich materials, characterized by high CaO content, can be identified or used for construction applications, while phosphate-rich zones, with elevated P&amp;#8322;O&amp;#8325; levels, can be flagged for potential recovery and further processing. This targeted approach ensures that waste materials are repurposed efficiently, aligning with circular economy principles. The study highlights the potential for automated, spectroscopy-based monitoring systems to support sustainable mining practices. Overall, this study demonstrates the power of combining cutting-edge remote sensing technologies with explainable machine learning to address the challenges of phosphate waste rock characterization. The methodology provides a scalable, cost-effective solution for mining operations worldwide, with significant implications for environmental sustainability, resource efficiency, and circular economy initiatives.Keywords:&amp;#160;Phosphate mine waste, Hyperspectral imaging, Field spectroscopy, Explainable machine learning (XML), Sustainable mining.

ObjectiveA rapid and accurate forecast for the early prognosis of ICH patients is challenging. This study investigated whether heart rate variability (HRV) and skin sympathetic nerve activity (SKNA) could prognosticate poor neurological outcomes in ICH patients.MethodsBetween November 2020 and November 2021, we studied 92 spontaneous ICH patients in the First Affiliated Hospital of Nanjing Medical University. Glasgow Outcome Scale (GOS) score at 2 weeks after the ICH was used to categorize patients into good and poor outcome groups. The modified Rankin Scale (mRS) assessed patients' ability to live independently for 1 year. We utilized a portable high‐frequency electrocardiogram (ECG) recording system to record the HRV and SKNA information in ICH patients and control participants.Results77 patients were eligible for the prediction of neurological outcome and were allocated into the good (n = 22) or poor (n = 55) outcome groups based on the GOS grade. In univariate logistic regression analysis, significant variables that could differentiate the outcomes were age, hypertension, tracheal intubation, Glasgow Coma Scale (GCS) score, existing intraventricular hemorrhage, white blood cells, neutrophil, lnVLF, lnTP, and aSKNA. Variables in the best fit multivariable logistic regression model were age, hypertension, GCS score, neutrophils, and aSKNA. The GCS score was the only independent risk factor for poor outcomes. At 30 days and 1 year of follow‐up, patients with lower aSKNA had poor outcomes.InterpretationICH patients had reduced aSKNA, which could be a prognostic indicator. A lower aSKNA suggested a worse prognosis. The present data indicate that ECG signals could be helpful for prognosticating ICH patients.

Trigeminal nerve stimulation for prolonged disorders of consciousness: A randomized double-blind sham-controlled study

Spatial-temporal patterns and drivers of coastal water quality dynamics: Insights from explainable machine learning and PLS-SEM analysis in Xiamen Bay, China.

Explainable machine learning using real, synthetic and augmented fire tests to predict fire resistance and spalling of RC columns

Electrical risk score and electrocardiographic markers in acute cholecystitis: a prospective cross-sectional study using explainable machine learning.