Dynamic prediction of mortality after traumatic brain injury using a machine learning algorithm

Artifact Management for Cerebral Near-Infrared Spectroscopy Signals: A Systematic Scoping Review

Artifacts induced during patient monitoring are a main limitation for near-infrared spectroscopy (NIRS) as a non-invasive method of cerebral hemodynamic monitoring. There currently does not exist a robust "gold-standard" method for artifact management for these signals. The objective of this review is to comprehensively examine the literature on existing artifact management methods for cerebral NIRS signals recorded in animals and humans. A search of five databases was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines. The search yielded 806 unique results. There were 19 articles from these results that were included in this review based on the inclusion/exclusion criteria. There were an additional 36 articles identified in the references of select articles that were also included. The methods outlined in these articles were grouped under two major categories: (1) motion and other disconnection artifact removal methods; (2) data quality improvement and physiological/other noise artifact filtering methods. These were sub-categorized by method type. It proved difficult to quantitatively compare the methods due to the heterogeneity of the effectiveness metrics and definitions of artifacts. The limitations evident in the existing literature justify the need for more comprehensive comparisons of artifact management. This review provides insights into the available methods for artifact management in cerebral NIRS and justification for a homogenous method to quantify the effectiveness of artifact management methods. This builds upon the work of two existing reviews that have been conducted on this topic; however, the scope is extended to all artifact types and all NIRS recording types. Future work by our lab in cerebral NIRS artifact management will lie in a layered artifact management method that will employ different techniques covered in this review (including dynamic thresholding, autoregressive-based methods, and wavelet-based methods) amongst others to remove varying artifact types.

Predictive Models of Long-Term Outcome in Patients with Moderate to Severe Traumatic Brain Injury are Biased Toward Mortality Prediction

BACKGROUND

The prognostication of long-term functional outcomes remains challenging in patients with traumatic brain injury (TBI). Our aim was to demonstrate that intensive care unit (ICU) variables are not efficient to predict 6-month functional outcome in survivors with moderate to severe TBI (msTBI) but are mostly associated with mortality, which leads to a mortality bias for models predicting a composite outcome of mortality and severe disability.

METHODS

We analyzed the data from the multicenter randomized controlled Continuous Hyperosmolar Therapy in Traumatic Brain-Injured Patients trial and developed predictive models using machine learning methods and baseline characteristics and predictors collected during ICU stay. We compared our models' predictions of 6-month binary Glasgow Outcome Scale extended (GOS-E) score in all patients with msTBI (unfavorable GOS-E 1-4 vs. favorable GOS-E 5-8) with mortality (GOS-E 1 vs. GOS-E 2-8) and binary functional outcome in survivors with msTBI (severe disability GOS-E 2-4 vs. moderate to no disability GOS-E 5-8). We investigated the link between ICU variables and long-term functional outcomes in survivors with msTBI using predictive modeling and factor analysis of mixed data and validated our hypotheses on the International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT) model.

RESULTS

Based on data from 370 patients with msTBI and classically used ICU variables, the prediction of the 6-month outcome in survivors was inefficient (mean area under the receiver operating characteristic 0.52). Using factor analysis of mixed data graph, we demonstrated that high-variance ICU variables were not associated with outcome in survivors with msTBI (p = 0.15 for dimension 1, p = 0.53 for dimension 2) but mostly with mortality (p < 0.001 for dimension 1), leading to a mortality bias for models predicting a composite outcome of mortality and severe disability. We finally identified this mortality bias in the IMPACT model.

CONCLUSIONS

We demonstrated using machine learning-based predictive models that classically used ICU variables are strongly associated with mortality but not with 6-month outcome in survivors with msTBI, leading to a mortality bias when predicting a composite outcome of mortality and severe disability.

Critical thresholds of long-pressure reactivity index and impact of intracranial pressure monitoring methods in traumatic brain injury

BACKGROUND

Moderate-to-severe traumatic brain injury (TBI) has a global mortality rate of about 30%, resulting in acquired life-long disabilities in many survivors. To potentially improve outcomes in this TBI population, the management of secondary injuries, particularly the failure of cerebrovascular reactivity (assessed via the pressure reactivity index; PRx, a correlation between intracranial pressure (ICP) and mean arterial blood pressure (MAP)), has gained interest in the field. However, derivation of PRx requires high-resolution data and expensive technological solutions, as calculations use a short time-window, which has resulted in it being used in only a handful of centers worldwide. As a solution to this, low resolution (longer time-windows) PRx has been suggested, known as Long-PRx or LPRx. Though LPRx has been proposed little is known about the best methodology to derive this measure, with different thresholds and time-windows proposed. Furthermore, the impact of ICP monitoring on cerebrovascular reactivity measures is poorly understood. Hence, this observational study establishes critical thresholds of LPRx associated with long-term functional outcome, comparing different time-windows for calculating LPRx as well as evaluating LPRx determined through external ventricular drains (EVD) vs intraparenchymal pressure device (IPD) ICP monitoring.

METHODS

The study included a total of n = 435 TBI patients from the Karolinska University Hospital. Patients were dichotomized into alive vs. dead and favorable vs. unfavorable outcomes based on 1-year Glasgow Outcome Scale (GOS). Pearson's chi-square values were computed for incrementally increasing LPRx or ICP thresholds against outcome. The thresholds that generated the greatest chi-squared value for each LPRx or ICP parameter had the highest outcome discriminatory capacity. This methodology was also completed for the segmentation of the population based on EVD, IPD, and time of data recorded in hospital stay.

RESULTS

LPRx calculated with 10-120-min windows behaved similarly, with maximal chi-square values ranging at around a LPRx of 0.25-0.35, for both survival and favorable outcome. When investigating the temporal relations of LPRx derived thresholds, the first 4 days appeared to be the most associated with outcomes. The segmentation of the data based on intracranial monitoring found limited differences between EVD and IPD, with similar LPRx values around 0.3.

CONCLUSION

Our work suggests that the underlying prognostic factors causing impairment in cerebrovascular reactivity can, to some degree, be detected using lower resolution PRx metrics (similar found thresholding values) with LPRx found clinically using as low as 10 min-by-minute samples of MAP and ICP. Furthermore, EVD derived LPRx with intermittent cerebrospinal fluid draining, seems to present similar outcome capacity as IPD. This low-resolution low sample LPRx method appears to be an adequate substitute for the clinical prognostic value of PRx and may be implemented independent of ICP monitoring method when PRx is not feasible, though further research is warranted.

Traumatic Brain Injury as an Independent Predictor of Futility in the Early Resuscitation of Patients in Hemorrhagic Shock

This review explores the concept of futility timeouts and the use of traumatic brain injury (TBI) as an independent predictor of the futility of resuscitation efforts in severely bleeding trauma patients. The national blood supply shortage has been exacerbated by the lingering influence of the COVID-19 pandemic on the number of blood donors available, as well as by the adoption of balanced hemostatic resuscitation protocols (such as the increasing use of 1:1:1 packed red blood cells, plasma, and platelets) with and without early whole blood resuscitation. This has underscored the urgent need for reliable predictors of futile resuscitation (FR). As a result, clinical, radiologic, and laboratory bedside markers have emerged which can accurately predict FR in patients with severe trauma-induced hemorrhage, such as the Suspension of Transfusion and Other Procedures (STOP) criteria. However, the STOP criteria do not include markers for TBI severity or transfusion cut points despite these patients requiring large quantities of blood components in the STOP criteria validation cohort. Yet, guidelines for neuroprognosticating patients with TBI can require up to 72 h, which makes them less useful in the minutes and hours following initial presentation. We examine the impact of TBI on bleeding trauma patients, with a focus on those with coagulopathies associated with TBI. This review categorizes TBI into isolated TBI (iTBI), hemorrhagic isolated TBI (hiTBI), and polytraumatic TBI (ptTBI). Through an analysis of bedside parameters (such as the proposed STOP criteria), coagulation assays, markers for TBI severity, and transfusion cut points as markers of futilty, we suggest amendments to current guidelines and the development of more precise algorithms that incorporate prognostic indicators of severe TBI as an independent parameter for the early prediction of FR so as to optimize blood product allocation.

Artificial Intelligence and Machine Learning Applications in Critically Ill Brain Injured Patients

The utilization of Artificial Intelligence (AI) and Machine Learning (ML) is paving the way for significant strides in patient diagnosis, treatment, and prognostication in neurocritical care. These technologies offer the potential to unravel complex patterns within vast datasets ranging from vast clinical data and EEG (electroencephalogram) readings to advanced cerebral imaging facilitating a more nuanced understanding of patient conditions. Despite their promise, the implementation of AI and ML faces substantial hurdles. Historical biases within training data, the challenge of interpreting multifaceted data streams, and the "black box" nature of ML algorithms present barriers to widespread clinical adoption. Moreover, ethical considerations around data privacy and the need for transparent, explainable models remain paramount to ensure trust and efficacy in clinical decision-making.This article reflects on the emergence of AI and ML as integral tools in neurocritical care, discussing their roles from the perspective of both their scientific promise and the associated challenges. We underscore the importance of extensive validation in diverse clinical settings to ensure the generalizability of ML models, particularly considering their potential to inform critical medical decisions such as withdrawal of life-sustaining therapies. Advancement in computational capabilities is essential for implementing ML in clinical settings, allowing for real-time analysis and decision support at the point of care. As AI and ML are poised to become commonplace in clinical practice, it is incumbent upon health care professionals to understand and oversee these technologies, ensuring they adhere to the highest safety standards and contribute to the realization of personalized medicine. This engagement will be pivotal in integrating AI and ML into patient care, optimizing outcomes in neurocritical care through informed and data-driven decision-making.

Compare the GCS and the Rotterdam CT Score in Predicting the Mortality and Disability of Patients with Traumatic Brain Injury

Background

Given the dearth of extensive research comparing the Glasgow Coma Scale with the Rotterdam scoring system for predicting mortality in trauma patients, this study was conducted to determine which scale provides a more realistic prediction of mortality in trauma patients after three months.

Materials and Methods

This observational study was performed at Kashani Hospital in Isfahan, Iran. Patients with TBI who were admitted between February 2022 and February 2023 were included in the study. Approval from the Ethical Committee of Isfahan University of Medical Sciences was obtained prior to conducting this study.

Results

We included 152 adult patients who completed the GOS-E and the QOLIBRI-OS three-month post-injury. The median age was 35 years (IQR = 17-70). Most patients 139 (91.4%) were classified as having a severe TBI.

Conclusion

The results of the present study showed that both the use of GCS and Rotterdam CT scores can be effective in predicting the three-month mortality and QOLIBRI-OS scores of patients, with the difference that the predictive power of the three-month Rotterdam CT score is greater than that of the GCS.

Refining outcome prediction after traumatic brain injury with machine learning algorithms

Outcome after traumatic brain injury (TBI) is typically assessed using the Glasgow outcome scale extended (GOSE) with levels from 1 (death) to 8 (upper good recovery). Outcome prediction has classically been dichotomized into either dead/alive or favorable/unfavorable outcome. Binary outcome prediction models limit the possibility of detecting subtle yet significant improvements. We set out to explore different machine learning methods with the purpose of mapping their predictions to the full 8 grade scale GOSE following TBI. The models were set up using the variables: age, GCS-motor score, pupillary reaction, and Marshall CT score. For model setup and internal validation, a total of 866 patients could be included. For external validation, a cohort of 369 patients were included from Leuven, Belgium, and a cohort of 573 patients from the US multi-center ProTECT III study. Our findings indicate that proportional odds logistic regression (POLR), random forest regression, and a neural network model achieved accuracy values of 0.3-0.35 when applied to internal data, compared to the random baseline which is 0.125 for eight categories. The models demonstrated satisfactory performance during external validation in the data from Leuven, however, their performance were not satisfactory when applied to the ProTECT III dataset.

Key Findings on Computed Tomography of the Head that Predict Death or the Need for Neurosurgical Intervention From Traumatic Brain Injury

BACKGROUND

Traumatic brain injury (TBI) requires rapid management to avoid secondary injury or death. This study evaluated if a simple schema for quickly interpreting CT head (CTH) imaging by trauma surgeons and trainees could be validated to predict need for neurosurgical intervention (NSI) or death from TBI within 24 hours.

METHODS

We retrospectively reviewed TBI patients presenting to our trauma center in 2020 with blunt mechanism and GCS ≤ 12. Primary independent variables were presence of 7 normal findings on CTH (CSF at foramen magnum, open fourth ventricle, CSF around quadrigeminal plate, CSF around cerebral peduncles, absence of midline shift, visible sulci/gyri, and gray-white differentiation). Trauma surgeons and trainees separately evaluated each patient's CTH, scoring findings as normal or abnormal. Primary outcome was NSI/death in 24 hours.

RESULTS

Our population consisted of 444 patients; 21.4% received NSI or died within 24 hours. By trainees' interpretation, 5.8% of patients without abnormal findings had NSI/death vs 52.0% of patients with ≥1 abnormality; attending interpretation was 8.7% and 54.9%, respectively (P < .001). Sulci/gyri effacement, midline shift, and cerebral peduncle effacement maximized sensitivity and specificity for predicting NSI/death. Considering pooled results, when ≥1 of those 3 findings was abnormal, sensitivity was 77.89%, specificity was 80.80%, positive predictive value was 52.48%, and negative predictive value was 93.07%.

DISCUSSION

Any single abnormality in this schema significantly predicted a large increase in NSI/death in 24 hours in TBI patients, and three particular findings were most predictive. This schema may help predict need for intervention and expedite management of moderate/severe TBI.

Development and validation of an interpretable Markov-embedded multilabel model for predicting risks of multiple postoperative complications among surgical inpatients: a multicenter prospective cohort study

BACKGROUND

When they encounter various highly related postoperative complications, existing risk evaluation tools that focus on single or any complications are inadequate in clinical practice. This seriously hinders complication management because of the lack of a quantitative basis. An interpretable multilabel model framework that predicts multiple complications simultaneously is urgently needed.

MATERIALS AND METHODS

The authors included 50 325 inpatients from a large multicenter cohort (2014-2017). The authors separated patients from one hospital for external validation and randomly split the remaining patients into training and internal validation sets. A MARKov-EmbeDded (MARKED) multilabel model was proposed, and three models were trained for comparison: binary relevance, a fully connected network (FULLNET), and a deep neural network. Performance was mainly evaluated using the area under the receiver operating characteristic curve (AUC). The authors interpreted the model using Shapley Additive Explanations. Complication-specific risk and risk source inference were provided at the individual level.

RESULTS

There were 26 292, 6574, and 17 459 inpatients in the training, internal validation, and external validation sets, respectively. For the external validation set, MARKED achieved the highest average AUC (0.818, 95% CI: 0.771-0.864) across eight outcomes [compared with binary relevance, 0.799 (0.748-0.849), FULLNET, 0.806 (0.756-0.856), and deep neural network, 0.815 (0.765-0.866)]. Specifically, the AUCs of MARKED were above 0.9 for cardiac complications [0.927 (0.894-0.960)], neurological complications [0.905 (0.870-0.941)], and mortality [0.902 (0.867-0.937)]. Serum albumin, surgical specialties, emergency case, American Society of Anesthesiologists score, age, and sex were the six most important preoperative variables. The interaction between complications contributed more than the preoperative variables, and formed a hierarchical chain of risk factors, mild complications, and severe complications.

CONCLUSION

The authors demonstrated the advantage of MARKED in terms of performance and interpretability. The authors expect that the identification of high-risk patients and the inference of the risk source for specific complications will be valuable for clinical decision-making.

Evaluation of Morlet Wavelet Analysis for Artifact Detection in Low-Frequency Commercial Near-Infrared Spectroscopy Systems

Regional cerebral oxygen saturation (rSO2), a method of cerebral tissue oxygenation measurement, is recorded using non-invasive near-infrared Spectroscopy (NIRS) devices. A major limitation is that recorded signals often contain artifacts. Manually removing these artifacts is both resource and time consuming. The objective was to evaluate the applicability of using wavelet analysis as an automated method for simple signal loss artifact clearance of rSO2 signals obtained from commercially available devices. A retrospective observational study using existing populations (healthy control (HC), elective spinal surgery patients (SP), and traumatic brain injury patients (TBI)) was conducted. Arterial blood pressure (ABP) and rSO2 data were collected in all patients. Wavelet analysis was determined to be successful in removing simple signal loss artifacts using wavelet coefficients and coherence to detect signal loss artifacts in rSO2 signals. The removal success rates in HC, SP, and TBI populations were 100%, 99.8%, and 99.7%, respectively (though it had limited precision in determining the exact point in time). Thus, wavelet analysis may prove to be useful in a layered approach NIRS signal artifact tool utilizing higher-frequency data; however, future work is needed.

Mortality prediction using medical time series on TBI patients

BACKGROUND AND OBJECTIVE

Traumatic Brain Injury (TBI) is one of the leading causes of injury-related mortality in the world, with severe cases reaching mortality rates of 30-40%. It is highly heterogeneous both in causes and consequences making more complex the medical interpretation and prognosis. Gathering clinical, demographic, and laboratory data to perform a prognosis requires time and skill in several clinical specialties. Artificial intelligence (AI) methods can take advantage of existing data by performing helpful predictions and guiding physicians toward a better prognosis and, consequently, better healthcare. The objective of this work was to develop learning models and evaluate their capability of predicting the mortality of TBI. The predictive model would allow the early assessment of the more serious cases and scarce medical resources can be pointed toward the patients who need them most.

METHODS

Long Short Term Memory (LSTM) and Transformer architectures were tested and compared in performance, coupled with data imbalance, missing data, and feature selection strategies. From the Medical Information Mart for Intensive Care III (MIMIC-III) dataset, a cohort of TBI patients was selected and an analysis of the first 48 hours of multiple time series sequential variables was done to predict hospital mortality.

RESULTS

The best performance was obtained with the Transformer architecture, achieving an AUC of 0.907 with the larger group of features and trained with class proportion class weights and binary cross entropy loss.

CONCLUSIONS

Using the time series sequential data, LSTM and Transformers proved to be both viable options for predicting TBI hospital mortality in 48 hours after admission. Overall, using sequential deep learning models with time series data to predict TBI mortality is viable and can be used as a helpful indicator of the well-being of patients.

Traumatic Brain Injury Rehabilitation Outcome Prediction Using Machine Learning Methods

Objective

To investigate the performance of machine learning (ML) methods for predicting outcomes from inpatient rehabilitation for subjects with TBI using a dataset with a large number of predictor variables. Our second objective was to identify top predictive features selected by the ML models for each outcome and to validate the interpretability of the models.

Design

Secondary analysis using computational modeling of relationships between patients, injury and treatment activities and 6 outcomes, applied to the large multi-site, prospective, longitudinal observational dataset collected during the traumatic brain injury inpatient rehabilitation study.

Setting

Acute inpatient rehabilitation.

Participants

1946 patients aged 14 years or older, who sustained a severe, moderate, or complicated mild TBI, and were admitted to 1 of 9 US inpatient rehabilitation sites between 2008 and 2011 (N=1946).

Main Outcome Measures

Rehabilitation length of stay, discharge to home, FIM cognitive and FIM motor at discharge and at 9-months post discharge.

Results

Advanced ML models, specifically gradient boosting tree model, performed consistently better than all other models, including classical linear regression models. Top ranked predictive features were identified for each of the 6 outcome variables. Level of effort, days to rehabilitation admission, age at rehabilitation admission, and advanced mobility activities were the most frequently top ranked predictive features. The highest-ranking predictive feature differed across the specific outcome variable.

Conclusions

Identifying patient, injury, and rehabilitation treatment variables that are predictive of better outcomes will contribute to cost-effective care delivery and guide evidence-based clinical practice. ML methods can contribute to these efforts.

Artificial intelligence in emergency medicine. A systematic literature review

Motivation and objective: Emergency medicine is becoming a popular application area for artificial intelligence methods but remains less investigated than other healthcare branches. The need for time-sensitive decision-making on the basis of high data volumes makes the use of quantitative technologies inevitable. However, the specifics of healthcare regulations impose strict requirements for such applications. Published contributions cover separate parts of emergency medicine and use disparate data and algorithms. This study aims to systematize the relevant contributions, investigate the main obstacles to artificial intelligence applications in emergency medicine, and propose directions for further studies.

METHODS

The contributions selection process was conducted with systematic electronic databases querying and filtering with respect to established exclusion criteria. Among the 380 papers gathered from IEEE Xplore, ACM Digital Library, Springer Library, ScienceDirect, and Nature databases 116 were considered to be a part of the survey. The main features of the selected papers are the focus on emergency medicine and the use of machine learning or deep learning algorithms.

FINDINGS AND DISCUSSION

The selected papers were classified into two branches: diagnostics-specific and triage-specific. The former ones are focused on either diagnosis prediction or decision support. The latter covers such applications as mortality, outcome, admission prediction, condition severity estimation, and urgent care prediction. The observed contributions are highly specialized within a single disease or medical operation and often use privately collected retrospective data, making them incomparable. These and other issues can be addressed by creating an end-to-end solution based on human-machine interaction.

CONCLUSION

Artificial intelligence applications are finding their place in emergency medicine, while most of the corresponding studies remain isolated and lack higher generalization and more sophisticated methodology, which can be a matter of forthcoming improvements.

Mining the contribution of intensive care clinical course to outcome after traumatic brain injury

Existing methods to characterise the evolving condition of traumatic brain injury (TBI) patients in the intensive care unit (ICU) do not capture the context necessary for individualising treatment. Here, we integrate all heterogenous data stored in medical records (1166 pre-ICU and ICU variables) to model the individualised contribution of clinical course to 6-month functional outcome on the Glasgow Outcome Scale -Extended (GOSE). On a prospective cohort (n = 1550, 65 centres) of TBI patients, we train recurrent neural network models to map a token-embedded time series representation of all variables (including missing values) to an ordinal GOSE prognosis every 2 h. The full range of variables explains up to 52% (95% CI: 50-54%) of the ordinal variance in functional outcome. Up to 91% (95% CI: 90-91%) of this explanation is derived from pre-ICU and admission information (i.e., static variables). Information collected in the ICU (i.e., dynamic variables) increases explanation (by up to 5% [95% CI: 4-6%]), though not enough to counter poorer overall performance in longer-stay (>5.75 days) patients. Highest-contributing variables include physician-based prognoses, CT features, and markers of neurological function. Whilst static information currently accounts for the majority of functional outcome explanation after TBI, data-driven analysis highlights investigative avenues to improve the dynamic characterisation of longer-stay patients. Moreover, our modelling strategy proves useful for converting large patient records into interpretable time series with missing data integration and minimal processing.

Machine Learning Approach for the Prediction of In-Hospital Mortality in Traumatic Brain Injury Using Bio-Clinical Markers at Presentation to the Emergency Department

BACKGROUND

Accurate prediction of in-hospital mortality is essential for better management of patients with traumatic brain injury (TBI). Machine learning (ML) algorithms have been shown to be effective in predicting clinical outcomes. This study aimed to identify predictors of in-hospital mortality in TBI patients using ML algorithms.

MATERIALS AND METHOD

A retrospective study was performed using data from both the trauma registry and electronic medical records among TBI patients admitted to the Hamad Trauma Center in Qatar between June 2016 and May 2021. Thirteen features were selected for four ML models including a Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (XgBoost), to predict the in-hospital mortality.

RESULTS

A dataset of 922 patients was analyzed, of which 78% survived and 22% died. The AUC scores for SVM, LR, XgBoost, and RF models were 0.86, 0.84, 0.85, and 0.86, respectively. XgBoost and RF had good AUC scores but exhibited significant differences in log loss between the training and testing sets (% difference in logloss of 79.5 and 41.8, respectively), indicating overfitting compared to the other models. The feature importance trend across all models indicates that aPTT, INR, ISS, prothrombin time, and lactic acid are the most important features in prediction. Magnesium also displayed significant importance in the prediction of mortality among serum electrolytes.

CONCLUSIONS

SVM was found to be the best-performing ML model in predicting the mortality of TBI patients. It had the highest AUC score and did not show overfitting, making it a more reliable model compared to LR, XgBoost, and RF.

Machine learning algorithms for improved prediction of in-hospital outcomes after moderate-to-severe traumatic brain injury: a Chinese retrospective cohort study

AIM

Controversy remains high over the superiority of advanced machine learning (ML) algorithms to conventional logistic regression (LR) in the prediction of prognosis after traumatic brain injury (TBI). This study aimed to compare the performance of ML and LR models in predicting in-hospital prognosis after TBI.

METHOD

In a single-center retrospective cohort of adult patients hospitalized for moderate-to-severe TBI (Glasgow coma score ≤12) in our hospital from 2011 to 2020, LR and three ML algorithms (XGboost, lightGBM, and FT-transformer) were run to build prediction models for in-hospital mortality and the Glasgow Outcome Scale (GOS) functional outcomes using either all 19 clinical and laboratory features or the 10 non-laboratory ones collected at admission to the neurological intensive care unit. The Shapley (SHAP) value was used for model interpretation.

RESULT

In total, 482 patients had an in-hospital mortality rate of 11.0%. A total of 23.0% of the patients had good functional scores (GOS ≥ 4) at discharge. All ML models performed better than the LR model in predicting in-hospital prognosis after TBI, among which the lightGBM model showed the best performance: When predicting mortality, the lightGBM model yielded an area under the curve (AUC) of 0.953 using all 19 features (the LR model: 0.813) and an AUC of 0.935 using 10 non-laboratory features (the LR model: 0.803); when predicting GOS functional outcomes, it yielded an AUC of 0.913 using all 19 features (the LR model: 0.832) and an AUC of 0.889 using non-laboratory data (the LR model: 0.818). The SHAP method identified key contributors to explain the lightGBM models. Finally, the integration of the lightGBM models with different prediction purposes was found to provide refined prognostic information, particularly for patients who survived moderate-to-severe TBI.

CONCLUSION

The study supported the superiority of ML to LR in predicting prognosis after moderate-to-severe TBI and highlighted its potential use for clinical application.

Prediction performance of the machine learning model in predicting mortality risk in patients with traumatic brain injuries: a systematic review and meta-analysis

PURPOSE

With the in-depth application of machine learning(ML) in clinical practice, it has been used to predict the mortality risk in patients with traumatic brain injuries(TBI). However, there are disputes over its predictive accuracy. Therefore, we implemented this systematic review and meta-analysis, to explore the predictive value of ML for TBI.

METHODOLOGY

We systematically retrieved literature published in PubMed, Embase.com, Cochrane, and Web of Science as of November 27, 2022. The prediction model risk of bias(ROB) assessment tool (PROBAST) was used to assess the ROB of models and the applicability of reviewed questions. The random-effects model was adopted for the meta-analysis of the C-index and accuracy of ML models, and a bivariate mixed-effects model for the meta-analysis of the sensitivity and specificity.

RESULT

A total of 47 papers were eligible, including 156 model, with 122 newly developed ML models and 34 clinically recommended mature tools. There were 98 ML models predicting the in-hospital mortality in patients with TBI; the pooled C-index, sensitivity, and specificity were 0.86 (95% CI: 0.84, 0.87), 0.79 (95% CI: 0.75, 0.82), and 0.89 (95% CI: 0.86, 0.92), respectively. There were 24 ML models predicting the out-of-hospital mortality; the pooled C-index, sensitivity, and specificity were 0.83 (95% CI: 0.81, 0.85), 0.74 (95% CI: 0.67, 0.81), and 0.75 (95% CI: 0.66, 0.82), respectively. According to multivariate analysis, GCS score, age, CT classification, pupil size/light reflex, glucose, and systolic blood pressure (SBP) exerted the greatest impact on the model performance.

CONCLUSION

According to the systematic review and meta-analysis, ML models are relatively accurate in predicting the mortality of TBI. A single model often outperforms traditional scoring tools, but the pooled accuracy of models is close to that of traditional scoring tools. The key factors related to model performance include the accepted clinical variables of TBI and the use of CT imaging.

Dietary (Poly)phenols in Traumatic Brain Injury

Traumatic brain injury (TBI) remains one of the leading causes of death and disability in young adults worldwide. Despite growing evidence and advances in our knowledge regarding the multifaceted pathophysiology of TBI, the underlying mechanisms, though, are still to be fully elucidated. Whereas initial brain insult involves acute and irreversible primary damage to the brain, the processes of subsequent secondary brain injury progress gradually over months to years, providing a window of opportunity for therapeutic interventions. To date, extensive research has been focused on the identification of druggable targets involved in these processes. Despite several decades of successful pre-clinical studies and very promising results, when transferred to clinics, these drugs showed, at best, modest beneficial effects, but more often, an absence of effects or even very harsh side effects in TBI patients. This reality has highlighted the need for novel approaches that will be able to respond to the complexity of the TBI and tackle TBI pathological processes on multiple levels. Recent evidence strongly indicates that nutritional interventions may provide a unique opportunity to enhance the repair processes after TBI. Dietary (poly)phenols, a big class of compounds abundantly found in fruits and vegetables, have emerged in the past few years as promising agents to be used in TBI settings due to their proven pleiotropic effects. Here, we give an overview of the pathophysiology of TBI and the underlying molecular mechanisms, followed by a state-of-the-art summary of the studies that have evaluated the efficacy of (poly)phenols administration to decrease TBI-associated damage in various animal TBI models and in a limited number of clinical trials. The current limitations on our knowledge concerning (poly)phenol effects in TBI in the pre-clinical studies are also discussed.