In-hospital mortality after traumatic brain injury surgery: a nationwide population-based comparison of mortality predictors used in artificial neural network and logistic regression models

Effect of low fibrinogen level on in-hospital mortality and 6-month functional outcome of TBI patients, a single center experience

In patients affected by traumatic brain injury (TBI), hypofibrinogenemia within the initial hours of trauma can be expected due to vascular and inflammatory changes. In this study, we aimed to evaluate the effect of hypofibrinogenemia on the in-hospital mortality and 6-month functional outcomes of TBI patients, admitted to Rajaee Hospital, a referral trauma center in Shiraz, Iran. This study included all TBI patients admitted to our center who had no prior history of coagulopathy or any systemic disease, were alive on arrival, and had not received any blood product before admission. On admission, hospitalization, imaging, and 6-month follow-up information of included patients were extracted from the TBI registry database. The baseline characteristics of patients with fibrinogen levels of less than 150 mg/dL were compared with the cases with higher levels. To assess the effect of low fibrinogen levels on in-hospital mortality, a uni- and multivariate was conducted between those who died in hospital and survivors. Based on the 6-month GOSE score of patients, those with GOSE < 4 (unfavorable outcome) were compared with those with a favorable outcome. A total of 3049 patients (84.3% male, 15.7% female), with a mean age of 39.25 ± 18.87, met the eligibility criteria of this study. 494 patients had fibrinogen levels < 150 mg/dl, who were mostly younger and had lower average GCS scores in comparison to cases with higher fibrinogen levels. By comparison of the patients who died during hospitalization and survivors, it was shown that fibrinogen < 150 mg/dl is among the prognostic factors for in-hospital mortality (OR:1.75, CI: 1.32:2.34, P-value < 0.001), while the comparison between patients with the favorable and unfavorable functional outcome at 6-month follow-up, was not in favor of prognostic effect of low fibrinogen level (OR: 0.80, CI: 0.58: 1.11, P-value: 0.19). Hypofibrinogenemia is associated with in-hospital mortality of TBI patients, along with known factors such as higher age and lower initial GCS score. However, it is not among the prognostic factors of midterm functional outcome.

The future of artificial intelligence in thoracic surgery for non-small cell lung cancer treatment a narrative review

Objectives

To present a comprehensive review of the current state of artificial intelligence (AI) applications in lung cancer management, spanning the preoperative, intraoperative, and postoperative phases.

Methods

A review of the literature was conducted using PubMed, EMBASE and Cochrane, including relevant studies between 2002 and 2023 to identify the latest research on artificial intelligence and lung cancer.

Conclusion

While AI holds promise in managing lung cancer, challenges exist. In the preoperative phase, AI can improve diagnostics and predict biomarkers, particularly in cases with limited biopsy materials. During surgery, AI provides real-time guidance. Postoperatively, AI assists in pathology assessment and predictive modeling. Challenges include interpretability issues, training limitations affecting model use and AI's ineffectiveness beyond classification. Overfitting and global generalization, along with high computational costs and ethical frameworks, pose hurdles. Addressing these challenges requires a careful approach, considering ethical, technical, and regulatory factors. Rigorous analysis, external validation, and a robust regulatory framework are crucial for responsible AI implementation in lung surgery, reflecting the evolving synergy between human expertise and technology.

An Artificial Intelligence Model for Predicting Trauma Mortality Among Emergency Department Patients in South Korea: Retrospective Cohort Study

BACKGROUND

Within the trauma system, the emergency department (ED) is the hospital's first contact and is vital for allocating medical resources. However, there is generally limited information about patients that die in the ED.

OBJECTIVE

The aim of this study was to develop an artificial intelligence (AI) model to predict trauma mortality and analyze pertinent mortality factors for all patients visiting the ED.

METHODS

We used the Korean National Emergency Department Information System (NEDIS) data set (N=6,536,306), incorporating over 400 hospitals between 2016 and 2019. We included the International Classification of Disease 10th Revision (ICD-10) codes and chose the following input features to predict ED patient mortality: age, sex, intentionality, injury, emergent symptom, Alert/Verbal/Painful/Unresponsive (AVPU) scale, Korean Triage and Acuity Scale (KTAS), and vital signs. We compared three different feature set performances for AI input: all features (n=921), ICD-10 features (n=878), and features excluding ICD-10 codes (n=43). We devised various machine learning models with an ensemble approach via 5-fold cross-validation and compared the performance of each model with that of traditional prediction models. Lastly, we investigated explainable AI feature effects and deployed our final AI model on a public website, providing access to our mortality prediction results among patients visiting the ED.

RESULTS

Our proposed AI model with the all-feature set achieved the highest area under the receiver operating characteristic curve (AUROC) of 0.9974 (adaptive boosting [AdaBoost], AdaBoost + light gradient boosting machine [LightGBM]: Ensemble), outperforming other state-of-the-art machine learning and traditional prediction models, including extreme gradient boosting (AUROC=0.9972), LightGBM (AUROC=0.9973), ICD-based injury severity scores (AUC=0.9328 for the inclusive model and AUROC=0.9567 for the exclusive model), and KTAS (AUROC=0.9405). In addition, our proposed AI model outperformed a cutting-edge AI model designed for in-hospital mortality prediction (AUROC=0.7675) for all ED visitors. From the AI model, we also discovered that age and unresponsiveness (coma) were the top two mortality predictors among patients visiting the ED, followed by oxygen saturation, multiple rib fractures (ICD-10 code S224), painful response (stupor, semicoma), and lumbar vertebra fracture (ICD-10 code S320).

CONCLUSIONS

Our proposed AI model exhibits remarkable accuracy in predicting ED mortality. Including the necessity for external validation, a large nationwide data set would provide a more accurate model and minimize overfitting. We anticipate that our AI-based risk calculator tool will substantially aid health care providers, particularly regarding triage and early diagnosis for trauma patients.

Machine Learning to Predict Three Types of Outcomes After Traumatic Brain Injury Using Data at Admission: A Multi-Center Study for Development and Validation

The difficulty of accurately identifying patients who would benefit from promising treatments makes it challenging to prove the efficacy of novel treatments for traumatic brain injury (TBI). Although machine learning is being increasingly applied to this task, existing binary outcome prediction models are insufficient for the effective stratification of TBI patients. The aim of this study was to develop an accurate 3-class outcome prediction model to enable appropriate patient stratification. To this end, retrospective balanced data of 1200 blunt TBI patients admitted to six Japanese hospitals from January 2018 onwards (200 consecutive cases at each institution) were used for model training and validation. We incorporated 21 predictors obtained in the emergency department, including age, sex, six clinical findings, four laboratory parameters, eight computed tomography findings, and an emergency craniotomy. We developed two machine learning models (XGBoost and dense neural network) and logistic regression models to predict 3-class outcomes based on the Glasgow Outcome Scale-Extended (GOSE) at discharge. The prediction models were developed using a training dataset with n = 1000, and their prediction performances were evaluated over two validation rounds on a validation dataset (n = 80) and a test dataset (n = 120) using the bootstrap method. Of the 1200 patients in aggregate, the median patient age was 71 years, 199 (16.7%) exhibited severe TBI, and emergency craniotomy was performed on 104 patients (8.7%). The median length of stay was 13.0 days. The 3-class outcomes were good recovery/moderate disability for 709 patients (59.1%), severe disability/vegetative state in 416 patients (34.7%), and death in 75 patients (6.2%). XGBoost model performed well with 69.5% sensitivity, 82.5% accuracy, and an area under the receiver operating characteristic curve of 0.901 in the final validation. In terms of the receiver operating characteristic curve analysis, the XGBoost outperformed the neural network-based and logistic regression models slightly. In particular, XGBoost outperformed the logistic regression model significantly in predicting severe disability/vegetative state. Although each model predicted favorable outcomes accurately, they tended to miss the mortality prediction. The proposed machine learning model was demonstrated to be capable of accurate prediction of in-hospital outcomes following TBI, even with the three GOSE-based categories. As a result, it is expected to be more impactful in the development of appropriate patient stratification methods in future TBI studies than conventional binary prognostic models. Further, outcomes were predicted based on only clinical data obtained from the emergency department. However, developing a robust model with consistent performance in diverse scenarios remains challenging, and further efforts are needed to improve generalization performance.

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.

Machine learning algorithms for predicting outcomes of traumatic brain injury: A systematic review and meta-analysis

BACKGROUND

Traumatic brain injury (TBI) is a leading cause of death and disability worldwide. The use of machine learning (ML) has emerged as a key advancement in TBI management. This study aimed to identify ML models with demonstrated effectiveness in predicting TBI outcomes.

METHODS

We conducted a systematic review in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis statement. In total, 15 articles were identified using the search strategy. Patient demographics, clinical status, ML outcome variables, and predictive characteristics were extracted. A small meta-analysis of mortality prediction was performed, and a meta-analysis of diagnostic accuracy was conducted for ML algorithms used across multiple studies.

RESULTS

ML algorithms including support vector machine (SVM), artificial neural networks (ANN), random forest, and Naïve Bayes were compared to logistic regression (LR). Thirteen studies found significant improvement in prognostic capability using ML versus LR. The accuracy of the above algorithms was consistently over 80% when predicting mortality and unfavorable outcome measured by Glasgow Outcome Scale. Receiver operating characteristic curves analyzing the sensitivity of ANN, SVM, decision tree, and LR demonstrated consistent findings across studies. Lower admission Glasgow Coma Scale (GCS), older age, elevated serum acid, and abnormal glucose were associated with increased adverse outcomes and had the most significant impact on ML algorithms.

CONCLUSION

ML algorithms were stronger than traditional regression models in predicting adverse outcomes. Admission GCS, age, and serum metabolites all have strong predictive power when used with ML and should be considered important components of TBI risk stratification.

Prediction prolonged mechanical ventilation in trauma patients of the intensive care unit according to initial medical factors: a machine learning approach

The goal of this study was to develop a predictive machine learning model to predict the risk of prolonged mechanical ventilation (PMV) in patients admitted to the intensive care unit (ICU), with a focus on laboratory and Arterial Blood Gas (ABG) data. This retrospective cohort study included ICU patients admitted to Rajaei Hospital in Shiraz between 2016 and March 20, 2022. All adult patients requiring mechanical ventilation and seeking ICU admission had their data analyzed. Six models were created in this study using five machine learning models (PMV more than 3, 5, 7, 10, 14, and 23 days). Patients' demographic characteristics, Apache II, laboratory information, ABG, and comorbidity were predictors. This study used Logistic regression (LR), artificial neural networks (ANN), support vector machines (SVM), random forest (RF), and C.5 decision tree (C.5 DT) to predict PMV. The study enrolled 1138 eligible patients, excluding brain-dead patients and those without mechanical ventilation or a tracheostomy. The model PMV > 14 days showed the best performance (Accuracy: 83.63-98.54). The essential ABG variables in our two optimal models (artificial neural network and decision tree) in the PMV > 14 models include FiO₂, paCO₂, and paO₂. This study provides evidence that machine learning methods outperform traditional methods and offer a perspective for achieving a consensus definition of PMV. It also introduces ABG and laboratory information as the two most important variables for predicting PMV. Therefore, there is significant value in deploying such models in clinical practice and making them accessible to clinicians to support their decision-making.

Prediction of motor function in patients with traumatic brain injury using genetic algorithms modified back propagation neural network: A data-based study

Objective

Traumatic brain injury (TBI) is one of the leading causes of death and disability worldwide. In this study, the characteristics of the patients, who were admitted to the China Rehabilitation Research Center, were elucidated in the TBI database, and a prediction model based on the Fugl-Meyer assessment scale (FMA) was established using this database.

Methods

A retrospective analysis of 463 TBI patients, who were hospitalized from June 2016 to June 2020, was performed. The data of the patients used for this study included the age and gender of the patients, course of TBI, complications, and concurrent dysfunctions, which were assessed using FMA and other measures. The information was collected at the time of admission to the hospital and 1 month after hospitalization. After 1 month, a prediction model, based on the correlation analyses and a 1-layer genetic algorithms modified back propagation (GA-BP) neural network with 175 patients, was established to predict the FMA. The correlations between the predicted and actual values of 58 patients (prediction set) were described.

Results

Most of the TBI patients, included in this study, had severe conditions (70%). The main causes of the TBI were car accidents (56.59%), while the most common complication and dysfunctions were hydrocephalus (46.44%) and cognitive and motor dysfunction (65.23 and 63.50%), respectively. A total of 233 patients were used in the prediction model, studying the 11 prognostic factors, such as gender, course of the disease, epilepsy, and hydrocephalus. The correlation between the predicted and the actual value of 58 patients was R ² = 0.95.

Conclusion

The genetic algorithms modified back propagation neural network can predict motor function in patients with traumatic brain injury, which can be used as a reference for risk and prognosis assessment and guide clinical decision-making.

Prognosis prediction in traumatic brain injury patients using machine learning algorithms

Predicting treatment outcomes in traumatic brain injury (TBI) patients is challenging worldwide. The present study aimed to achieve the most accurate machine learning (ML) algorithms to predict the outcomes of TBI treatment by evaluating demographic features, laboratory data, imaging indices, and clinical features. We used data from 3347 patients admitted to a tertiary trauma centre in Iran from 2016 to 2021. After the exclusion of incomplete data, 1653 patients remained. We used ML algorithms such as random forest (RF) and decision tree (DT) with ten-fold cross-validation to develop the best prediction model. Our findings reveal that among different variables included in this study, the motor component of the Glasgow coma scale, the condition of pupils, and the condition of cisterns were the most reliable features for predicting in-hospital mortality, while the patients' age takes the place of cisterns condition when considering the long-term survival of TBI patients. Also, we found that the RF algorithm is the best model to predict the short-term mortality of TBI patients. However, the generalized linear model (GLM) algorithm showed the best performance (with an accuracy rate of 82.03 ± 2.34) in predicting the long-term survival of patients. Our results showed that using appropriate markers and with further development, ML has the potential to predict TBI patients' survival in the short- and long-term.

Machine Learning in the Prediction of Trauma Outcomes: A Systematic Review

STUDY OBJECTIVE

Machine learning models carry unique potential as decision-making aids and prediction tools for improving patient care. Traumatically injured patients provide a uniquely heterogeneous population with severe injuries that can be difficult to predict. Given the relative infancy of machine learning applications in medicine, this systematic review aimed to better understand the current state of machine learning development and implementation to help create a basis for future research.

METHODS

We conducted a systematic review from inception to May 2021, using Embase, MEDLINE through Ovid, Web of Science, Google Scholar, and relevant gray literature, for uses of machine learning in predicting the outcomes of trauma patients. The screening and data extraction were performed by 2 independent reviewers.

RESULTS

Of the 14,694 identified articles screened, 67 were included for data extraction. Artificial neural networks comprised the most commonly used model, and mortality was the most prevalent outcome of interest. In terms of machine learning model development, there was a lack of studies that employed external validation, feature selection methods, and performed formal calibration testing. Significant heterogeneity in reporting was also observed between the machine learning models employed, patient populations, performance metrics, and features employed.

CONCLUSION

This review highlights the heterogeneity in the development and reporting of machine learning models for the prediction of trauma outcomes. While these models present an area of opportunity as an ancillary to clinical decision-making, we recommend more standardization and rigorous guidelines for the development of future models.

A Computer-Assisted System for Early Mortality Risk Prediction in Patients with Traumatic Brain Injury Using Artificial Intelligence Algorithms in Emergency Room Triage

Traumatic brain injury (TBI) remains a critical public health challenge. Although studies have found several prognostic factors for TBI, a useful early predictive tool for mortality has yet to be developed in the triage of the emergency room. This study aimed to use machine learning algorithms of artificial intelligence (AI) to develop predictive models for TBI patients in the emergency room triage. We retrospectively enrolled 18,249 adult TBI patients in the electronic medical records of three hospitals of Chi Mei Medical Group from January 2010 to December 2019, and undertook the 12 potentially predictive feature variables for predicting mortality during hospitalization. Six machine learning algorithms including logistical regression (LR) random forest (RF), support vector machines (SVM), LightGBM, XGBoost, and multilayer perceptron (MLP) were used to build the predictive model. The results showed that all six predictive models had high AUC from 0.851 to 0.925. Among these models, the LR-based model was the best model for mortality risk prediction with the highest AUC of 0.925; thus, we integrated the best model into the existed hospital information system for assisting clinical decision-making. These results revealed that the LR-based model was the best model to predict the mortality risk in patients with TBI in the emergency room. Since the developed prediction system can easily obtain the 12 feature variables during the initial triage, it can provide quick and early mortality prediction to clinicians for guiding deciding further treatment as well as helping explain the patient’s condition to family members.

Evaluation of machine learning algorithms for trabeculectomy outcome prediction in patients with glaucoma

The purpose of this study was to evaluate the performance of machine learning algorithms to predict trabeculectomy surgical outcomes. Preoperative systemic, demographic and ocular data from consecutive trabeculectomy surgeries from a single academic institution between January 2014 and December 2018 were incorporated into models using random forest, support vector machine, artificial neural networks and multivariable logistic regression. Mean area under the receiver operating characteristic curve (AUC) and accuracy were used to evaluate the discrimination of each model to predict complete success of trabeculectomy surgery at 1 year. The top performing model was optimized using recursive feature selection and hyperparameter tuning. Calibration and net benefit of the final models were assessed. Among the 230 trabeculectomy surgeries performed on 184 patients, 104 (45.2%) were classified as complete success. Random forest was found to be the top performing model with an accuracy of 0.68 and AUC of 0.74 using 5-fold cross-validation to evaluate the final optimized model. These results provide evidence that machine learning models offer value in predicting trabeculectomy outcomes in patients with refractory glaucoma.

A machine learning model for predicting favorable outcome in severe traumatic brain injury patients after 6 months

Traumatic brain injury (TBI), which occurs commonly worldwide, is among the more costly of health and socioeconomic problems. Accurate prediction of favorable outcome in severe TBI patients could assist with optimizing treatment procedures, predicting clinical outcomes, and result in substantial economic savings. In this study, we examined the capability of a machine learning-based model in predicting "favorable" or "unfavorable" outcome after 6 months in severe TBI patients using only parameters measured on admission. Three models were developed using logistic regression, random forest, and support vector machines trained on parameters recorded from 2,381 severe TBI patients admitted to the neuro-intensive care unit of Rajaee (Emtiaz) Hospital (Shiraz, Iran) between 2015 and 2017. Model performance was evaluated using three indices: sensitivity, specificity, accuracy, and area under the curve (AUC). Ten-fold cross-validation method was used to estimate these indices. Overall, the developed models showed excellent performance with AUC >0.81, sensitivity and specificity of > 0.78. The top-three factors important in predicting 6-month post-trauma survival status in TBI patients are "GCS motor response," "pupillary reactivity," and "age." Machine learning techniques might be used to predict the 6-month outcome in TBI patients using only the parameters measured on admission when the machine learning is trained using a large data set.

Which criteria is a better predictor of ICU admission in trauma patients? An artificial neural network approach

PURPOSE

One of the most critical concerns in the intensive care unit (ICU) section is identifying the best criteria for entering patients to this part. This study aimed to predict the best compatible criteria for entering trauma patients in the ICU section.

METHOD

The present study was a historical cohort study. The data were collected from 2448 trauma patients referring to Shahid Rajaee Hospital between January 2015 and January 2017 in Shiraz, Iran. The artificial neural network (ANN) models with cross-validation and logistic regression (LR) with a backward method was used for data analysis. The final analysis was performed on a total of 958 patients who were transferred to the ICU section.

RESULTS

Based on the present results, the motor component of the GCS score at each cutoff point had the highest importance. The results also showed better performance for the AUC and accuracy rate for ANN compared with LR.

CONCLUSION

The most critical indicators in predicting the optimal use of ICU services in this study were the Motor component of the GCS. Results revealed that the ANN had a better performance than the LR in predicting the main outcomes of the traumatic patients in both the accuracy and AUC index. Trauma section surgeons and ICU specialists will benefit from this study's results and can assist them in making decisions to predict the patient outcomes before entering the ICU.

The detection of mild traumatic brain injury in paediatrics using artificial neural networks

Head computed tomography (CT) is the gold standard in emergency departments (EDs) to evaluate mild traumatic brain injury (mTBI) patients, especially for paediatrics. Data-driven models for successfully classifying head CT scans that have mTBI will be valuable in terms of timeliness and cost-effectiveness for TBI diagnosis. This study applied two different machine learning (ML) models to diagnose mTBI in a paediatric population collected as part of the paediatric emergency care applied research network (PECARN) study between 2004 and 2006. The models were conducted using 15,271 patients under the age of 18 years with mTBI and had a head CT report. In the conventional model, random forest (RF) ranked the features to reduce data dimensionality and the top ranked features were used to train a shallow artificial neural network (ANN) model. In the second model, a deep ANN applied to classify positive and negative mTBI patients using the entirety of the features available. The dataset was divided into two subsets: 80% for training and 20% for testing using five-fold cross-validation. Accuracy, sensitivity, precision, and specificity were calculated by comparing the model's prediction outcome to the actual diagnosis for each patient. RF ranked ten clinical demographic features and twelve CT-findings; the hybrid RF-ANN model achieved an average specificity of 99.96%, sensitivity of 95.98%, precision of 99.25%, and accuracy of 99.74% in identifying positive mTBI from negative mTBI subjects. The deep ANN proved its ability to carry out the task efficiently with an average specificity of 99.9%, sensitivity of 99.2%, precision of 99.9%, and accuracy of 99.9%. The performance of the two proposed models demonstrated the feasibility of using ANN to diagnose mTBI in a paediatric population. This is the first study to investigate deep ANN in a paediatric cohort with mTBI using clinical and non-imaging data and diagnose mTBI with balanced sensitivity and specificity using shallow and deep ML models. This method, if validated, would have the potential to reduce the burden of TBI evaluation in EDs and aide clinicians in the decision-making process.

Heart Rate Variability as a Predictor of Mortality in Traumatic Brain Injury: A Systematic Review and Meta-Analysis

OBJECTIVE

To systematically review the medical literature to determine the utility of heart rate variability in predicting mortality for moderate to severe traumatic brain injury.

METHODS

A search for randomized controlled trials, nonrandomized trials, and prospective and retrospective cohort studies was carried out using PubMed, SCOPUS, Cochrane Central Register of Controlled Trials, MEDLINE, and EMBASE. Reference lists of included studies were also searched to identify potentially eligible studies.

RESULTS

Five articles comprising 542 patients met inclusion criteria. Heart rate variability as low-frequency/high-frequency ratio (area under the curve [AUC] receiver operating characteristic [ROC]) for predicting mortality was found to be statistically significant (AUC ROC 0.810, P < 0.001) with high heterogeneity (I² = 61.98%, P = 0.032). Meta-analysis of low-frequency/high-frequency ratio, High frequency peak, and total power were statistically significant for predicting mortality. Odd's ratio for predicting mortality for LF/HF ratio, HF peak, and TP were 16.17, 19.09, 22.59 respectively. High-frequency peak in predicting mortality showed an AUC ROC of 0.986 (P ≤ 0.001) with a low level of heterogeneity. Total power (TP) showed an AUC ROC of 0.93 (P < 0.001) in predicting mortality with a high level of heterogeneity (I² = 83.16%, P = 0.002). Funnel plot analysis to assess the presence of publication bias for TP showed a high level of heterogeneity and asymmetry among studies.

CONCLUSIONS

This meta-analysis predicted high mortality based on odds ratio for variables low-frequency/high-frequency ratio, high-frequency peak, and TP. However, the statistical analysis was weakened owing to the high level of heterogeneity in the included studies. Further research is needed to generate high-quality recommendations regarding heart rate variability as a predictor of mortality after traumatic brain injury.

Predicting the Individual Treatment Effect of Neurosurgery for Patients with Traumatic Brain Injury in the Low-Resource Setting: A Machine Learning Approach in Uganda

Traumatic brain injury (TBI) disproportionately affects low- and middle-income countries (LMICs). In these low-resource settings, effective triage of patients with TBI-including the decision of whether or not to perform neurosurgery-is critical in optimizing patient outcomes and healthcare resource utilization. Machine learning may allow for effective predictions of patient outcomes both with and without surgery. Data from patients with TBI was collected prospectively at Mulago National Referral Hospital in Kampala, Uganda, from 2016 to 2019. One linear and six non-linear machine learning models were designed to predict good versus poor outcome near hospital discharge and internally validated using nested five-fold cross-validation. The 13 predictors included clinical variables easily acquired on admission and whether or not the patient received surgery. Using an elastic-net regularized logistic regression model (GLMnet), with predictions calibrated using Platt scaling, the probability of poor outcome was calculated for each patient both with and without surgery (with the difference quantifying the "individual treatment effect," ITE). Relative ITE represents the percent reduction in chance of poor outcome, equaling this ITE divided by the probability of poor outcome with no surgery. Ultimately, 1766 patients were included. Areas under the receiver operating characteristic curve (AUROCs) ranged from 83.1% (single C5.0 ruleset) to 88.5% (random forest), with the GLMnet at 87.5%. The two variables promoting good outcomes in the GLMnet model were high Glasgow Coma Scale score and receiving surgery. For the subgroup not receiving surgery, the median relative ITE was 42.9% (interquartile range [IQR], 32.7% to 53.5%); similarly, in those receiving surgery, it was 43.2% (IQR, 32.9% to 54.3%). We provide the first machine learning-based model to predict TBI outcomes with and without surgery in LMICs, thus enabling more effective surgical decision making in the resource-limited setting. Predicted ITE similarity between surgical and non-surgical groups suggests that, currently, patients are not being chosen optimally for neurosurgical intervention. Our clinical decision aid has the potential to improve outcomes.

[Artificial intelligence in neurosurgery]

BACKGROUND

Artificial intelligence (AI) in neurosurgery is becoming increasingly more important as the technology advances. This development can be measured by the increase of publications on AI in neurosurgery over the last years.

OBJECTIVE

This article provides insights into the current possibilities of using AI in neurosurgery.

MATERIAL AND METHODS

A review of the literature was carried out with a focus on exemplary work on the use of AI in neurosurgery.

RESULTS

The current neurosurgical publications on the use of AI show the diversity of the topic in this field. The main areas of application are diagnostics, outcome and treatment models.

CONCLUSION

The various areas of application of AI in the field of neurosurgery with a refined preoperative diagnostics and outcome predictions will significantly influence the future of neurosurgery. Neurosurgeons will continue to make the decisions on the indications for surgery but an optimized statement on diagnosis, treatment options and on the risk of surgery will be made by neurosurgeons with the help of AI in the future.

Using trauma registry data to predict prolonged mechanical ventilation in patients with traumatic brain injury: Machine learning approach

OBJECTIVES

We aimed to build a machine learning predictive model to predict the risk of prolonged mechanical ventilation (PMV) for patients with Traumatic Brain Injury (TBI).

METHODS

This study included TBI patients who were hospitalized in a level 1 trauma center between January 2014 and February 2019. Data were analyzed for all adult patients who received mechanical ventilation following TBI with abbreviated injury severity (AIS) score for the head region of ≥ 3. This study designed three sets of machine learning models: set A defined PMV to be greater than 7 days, set B (PMV > 10 days) and set C (PMV >14 days) to determine the optimal model for deployment. Patients' demographics, injury characteristics and CT findings were used as predictors. Logistic regression (LR), Artificial neural networks (ANN) Support vector machines (SVM), Random Forest (RF) and C.5 Decision Tree (C.5 DT) were used to predict the PMV.

RESULTS

The number of eligible patients that were included in the study were 674, 643 and 622 patients in sets A, B and C respectively. In set A, LR achieved the optimal performance with accuracy 0.75 and Area under the curve (AUC) 0.83. SVM achieved the optimal performance among other models in sets B with accuracy/AUC of 0.79/0.84 respectively. ANNs achieved the optimal performance in set C with accuracy/AUC of 0.76/0.72 respectively. Machine learning models in set B demonstrated more stable performance with higher prediction success and discrimination power.

CONCLUSION

This study not only provides evidence that machine learning methods outperform the traditional multivariate analytical methods, but also provides a perspective to reach a consensual definition of PMV.

A traumatic brain injury prognostic model to support in-hospital triage in a low-income country: a machine learning-based approach

OBJECTIVE

Traumatic brain injury (TBI) is a leading cause of death and disability worldwide, with a disproportionate burden of this injury on low- and middle-income countries (LMICs). Limited access to diagnostic technologies and highly skilled providers combined with high patient volumes contributes to poor outcomes in LMICs. Prognostic modeling as a clinical decision support tool, in theory, could optimize the use of existing resources and support timely treatment decisions in LMICs. The objective of this study was to develop a machine learning-based prognostic model using data from Kilimanjaro Christian Medical Centre in Moshi, Tanzania.

METHODS

This study is a secondary analysis of a TBI data registry including 3138 patients. The authors tested nine different machine learning techniques to identify the prognostic model with the greatest area under the receiver operating characteristic curve (AUC). Input data included demographics, vital signs, injury type, and treatment received. The outcome variable was the discharge score on the Glasgow Outcome Scale-Extended.

RESULTS

The AUC for the prognostic models varied from 66.2% (k-nearest neighbors) to 86.5% (Bayesian generalized linear model). An increasing Glasgow Coma Scale score, increasing pulse oximetry values, and undergoing TBI surgery were predictive of a good recovery, while injuries suffered from a motor vehicle crash and increasing age were predictive of a poor recovery.

CONCLUSIONS

The authors developed a TBI prognostic model with a substantial level of accuracy in a low-resource setting. Further research is needed to externally validate the model and test the algorithm as a clinical decision support tool.

The applications of machine learning in plastic and reconstructive surgery: protocol of a systematic review

BACKGROUND

Machine learning, a subset of artificial intelligence, is a set of models and methods that can automatically detect patterns in vast amounts of data, extract information and use it to perform various kinds of decision-making under uncertain conditions. This can assist surgeons in clinical decision-making by identifying patient cohorts that will benefit from surgery prior to treatment. The aim of this review is to evaluate the applications of machine learning in plastic and reconstructive surgery.

METHODS

A literature review will be undertaken of EMBASE, MEDLINE and CENTRAL (1990 up to September 2019) to identify studies relevant for the review. Studies in which machine learning has been employed in the clinical setting of plastic surgery will be included. Primary outcomes will be the evaluation of the accuracy of machine learning models in predicting a clinical diagnosis and post-surgical outcomes. Secondary outcomes will include a cost analysis of those models. This protocol has been prepared using the Preferred Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guidelines.

DISCUSSION

This will be the first systematic review in available literature that summarises the published work on the applications of machine learning in plastic surgery. Our findings will provide the basis of future research in developing artificial intelligence interventions in the specialty.

SYSTEMATIC REVIEW REGISTRATION

PROSPERO CRD42019140924.

Development of an unsupervised machine learning algorithm for the prognostication of walking ability in spinal cord injury patients

BACKGROUND CONTEXT

Traumatic spinal cord injury can have a dramatic effect on a patient's life. The degree of neurologic recovery greatly influences a patient's treatment and expected quality of life. This has resulted in the development of machine learning algorithms (MLA) that use acute demographic and neurologic information to prognosticate recovery. The van Middendorp et al. (2011) (vM) logistic regression (LR) model has been established as a reference model for the prediction of walking recovery following spinal cord injury as it has been validated within many different countries. However, an examination of the way in which these prediction models are evaluated is warranted. The area under the receiver operators curve (AUROC) has been consistently used when evaluating model performance, but it has been shown that AUROC overemphasizes the most common event resulting in an inaccurate assessment when the data are imbalanced. Furthermore, there is evidence that the use of more advanced MLA, such as an unsupervised k-means model, may show superior performance compared to LR as they can handle a larger number of features.

PURPOSE

The first objective of the study was to assess the performance of both an unsupervised MLA and LR model with complete admission neurologic information against the vM and Hicks models. Second, a comparison between the accuracy of the AUROC and the F1-score will be made to determine which method is superior for the assessment of diagnostic performance of prediction models on large-scale datasets.

STUDY DESIGN

Retrospective review of a prospective cohort study.

PATIENT SAMPLE

The Rick Hansen Spinal Cord Injury Registry (RHSCIR) was used in this study. All patients enrolled between 2004 and 2017 with complete neurologic examination and Functional Independence Measure outcome data at ≥1 year follow-up or who could walk at discharge were included. The prognostic variables included age (dichotomized at ≥65 years old); American Spinal Injury Association Impairment Scale (AIS) grade; and individual motor, light touch, and pinprick score from L2 to S1.

OUTCOME MEASURES

The Functional Independence Measure locomotor score was used to assess independent walking ability at discharge or 1-year follow-up.

METHODS

An unsupervised MLA with k=2 was chosen in order to identify a "walk" cluster and a "not walk" cluster. Model performance was assessed through the development of a receiver operating characteristic curve with associated AUROC and a precision-recall curve with associated F1-score. The study and the RHSCIR are supported by funding from Health Canada, Western Economic Diversification Canada, and the Governments of Alberta, British Columbia, Manitoba, and Ontario. These funders had no role in the study or study reporting and the authors have no conflicts of interest to report.

RESULTS

No clinically relevant differences were found between with the use of an unsupervised MLA with a greater amount of initial neurologic information compared to the established standards for any AIS classification. Although demonstrated for all separate AIS classifications, most notably, the AUROC for the vM (0.78) and Hicks models (0.76) were found to be superior to that of the new LR model (0.72); however, the vM and Hicks models had more than double the amount of false negative classifications compared to the LR. The F1-scores between these three models were also found to be different but with the vM and Hicks models being lower than the LR (0.85, 0.81, and 0.89, respectively).

CONCLUSIONS

No clinically relevant differences were found between the use of an unsupervised MLA with complete admission neurologic information compared to the previously validated standards; however, when comparing the performance of the AUROC and F1-score, the AUROC showed inaccurate prognostic performance when there was an imbalance toward a greater amount of false negatives. Importantly, the F1-score did not succumb to this imbalance. As AUROC has been used as the standard when evaluating performance of prediction models, consideration as to whether this is the most appropriate method is warranted. Future work should focus on comparing AUROC and F1-scores with other previously validated models.

Predictive models for patients with lung carcinomas to identify EGFR mutation status via an artificial neural network based on multiple clinical information

PURPOSE

Epidermal growth factor receptor (EGFR) mutation testing has several limitations. Therefore, we built predictive models to determine the EGFR mutation status of patients and guide therapeutic decision-making.

METHODS

We collected data from 320 patients with lung carcinoma, including sex, age, smoking history, serum tumour marker levels, maximum standardized uptake value, pathological results, computed tomography images, and EGFR mutation status. Artificial neural network (ANN) models based on multiple clinical characteristics were proposed to predict EGFR mutation status.

RESULTS

A training set (n = 200) was used to develop predictive models of the EGFR mutation status (Model 1: area under the receiver operating characteristic curve [AUROC] = 0.910, 95% CI 0.861-0.945; Model 2: AUROC = 0.859, 95% CI 0.803-0.904; Model 3: AUROC = 0.711, 95% CI 0.643-0.773). A testing set (n = 50) and temporal validation data set (n = 70) were used to evaluate the generalisation performance of the established models (testing set: Model 1, AUROC = 0.845, 95% CI 0.715-0.932; Model 2, AUROC = 0.882, 95% CI 0.759-0.956; Model 3, AUROC = 0.817, 95% CI 0.682-0.912; temporal validation dataset: Model 1, AUROC = 0.909, 95% CI 0.816-0.964; Model 2, AUROC = 0.855, 95% CI 0.751-0.928; Model 3, AUROC = 0.831, 95% CI 0.723-0.910). The predictive abilities of the three ANN models were superior to that of a previous logistic regression model (P < 0.001, 0.027, and 0.050, respectively).

CONCLUSIONS

ANN models provide a non-invasive and readily available method for EGFR mutation status prediction.

Identifying Ear Abnormality from 2D Photographs Using Convolutional Neural Networks

Quantifying ear deformity using linear measurements and mathematical modeling is difficult due to the ear's complex shape. Machine learning techniques, such as convolutional neural networks (CNNs), are well-suited for this role. CNNs are deep learning methods capable of finding complex patterns from medical images, automatically building solution models capable of machine diagnosis. In this study, we applied CNN to automatically identify ear deformity from 2D photographs. Institutional review board (IRB) approval was obtained for this retrospective study to train and test the CNNs. Photographs of patients with and without ear deformity were obtained as standard of care in our photography studio. Profile photographs were obtained for one or both ears. A total of 671 profile pictures were used in this study including: 457 photographs of patients with ear deformity and 214 photographs of patients with normal ears. Photographs were cropped to the ear boundary and randomly divided into training (60%), validation (20%), and testing (20%) datasets. We modified the softmax classifier in the last layer in GoogLeNet, a deep CNN, to generate an ear deformity detection model in Matlab. All images were deemed of high quality and usable for training and testing. It took about 2 hours to train the system and the training accuracy reached almost 100%. The test accuracy was about 94.1%. We demonstrate that deep learning has a great potential in identifying ear deformity. These machine learning techniques hold the promise in being used in the future to evaluate treatment outcomes.

Application of three statistical models for predicting the risk of diabetes

BACKGROUND

At present, the proportion of undiagnosed diabetes in Chinese adults is as high as 15.5%. People with diabetes who are not treated and controlled in time may have various complications, such as cardiovascular and cerebrovascular diseases and diabetic foot disorders, which not only seriously affect the quality of life of people with diabetes but also impose a heavy burden on families and society. Therefore, prevention and control of type 2 diabetes is of great significance.

METHODS

We constructed a logistic regression model, a neural network model and a decision tree model to analyse the risk factors for type 2 diabetes and then compared the prediction accuracy of the different models by calculating the area under the relative operating characteristic (ROC) curve and back-inputting the data into the model.

RESULTS

The prevalence of type 2 diabetes in 4177 subjects who were not diagnosed with type 2 diabetes was 9.31%. The most influential factors associated with type 2 diabetes were triglyceride (TG) ≥ 1.17 mmol/L (odds ratio (OR) =2.233), age ≥ 70 years (OR = 1.734), hypertension (OR = 1.703), alcohol consumption (OR = 1.674), and total cholesterol≥5.2 mmol/L (TC) (OR = 1.463). The prediction accuracies of the three prediction models were 90.8, 91.2, and 90.7%, respectively, and the areas under curve (AUCs) were 0.711, 0.780, and 0.698, respectively. The differences in the AUCs after back propagation (BP) of the neural network model, logistic regression model and decision tree model were statistically significant (P < 0.05).

CONCLUSION

BP neural networks have a higher predictive power for identifying the associated risk factors of type 2 diabetes than the other two models, but it is necessary to select a suitable model for specific situations.

Machine-learning analysis outperforms conventional statistical models and CT classification systems in predicting 6-month outcomes in pediatric patients sustaining traumatic brain injury

OBJECTIVEModern surgical planning and prognostication requires the most accurate outcomes data to practice evidence-based medicine. For clinicians treating children following traumatic brain injury (TBI) these data are severely lacking. The first aim of this study was to assess published CT classification systems in the authors' pediatric cohort. A pediatric-specific machine-learning algorithm called an artificial neural network (ANN) was then created that robustly outperformed traditional CT classification systems in predicting TBI outcomes in children.METHODSThe clinical records of children under the age of 18 who suffered a TBI and underwent head CT within 24 hours after TBI (n = 565) were retrospectively reviewed.RESULTS"Favorable" outcome (alive with Glasgow Outcome Scale [GOS] score ≥ 4 at 6 months postinjury, n = 533) and "unfavorable" outcome (death at 6 months or GOS score ≤ 3 at 6 months postinjury, n = 32) were used as the primary outcomes. The area under the receiver operating characteristic (ROC) curve (AUC) was used to delineate the strength of each CT grading system in predicting survival (Helsinki, 0.814; Rotterdam, 0.838; and Marshall, 0.781). The AUC for CT score in predicting GOS score ≤ 3, a measure of overall functionality, was similarly predictive (Helsinki, 0.717; Rotterdam, 0.748; and Marshall, 0.663). An ANN was then constructed that was able to predict 6-month outcomes with profound accuracy (AUC = 0.9462 ± 0.0422).CONCLUSIONSThis study showed that machine-learning can be leveraged to more accurately predict TBI outcomes in children.

Machine learning analyses can differentiate meningioma grade by features on magnetic resonance imaging

OBJECTIVEPrognostication and surgical planning for WHO grade I versus grade II meningioma requires thoughtful decision-making based on radiographic evidence, among other factors. Although conventional statistical models such as logistic regression are useful, machine learning (ML) algorithms are often more predictive, have higher discriminative ability, and can learn from new data. The authors used conventional statistical models and an array of ML algorithms to predict atypical meningioma based on radiologist-interpreted preoperative MRI findings. The goal of this study was to compare the performance of ML algorithms to standard statistical methods when predicting meningioma grade.METHODSThe cohort included patients aged 18-65 years with WHO grade I (n = 94) and II (n = 34) meningioma in whom preoperative MRI was obtained between 1998 and 2010. A board-certified neuroradiologist, blinded to histological grade, interpreted all MR images for tumor volume, degree of peritumoral edema, presence of necrosis, tumor location, presence of a draining vein, and patient sex. The authors trained and validated several binary classifiers: k-nearest neighbors models, support vector machines, naïve Bayes classifiers, and artificial neural networks as well as logistic regression models to predict tumor grade. The area under the curve-receiver operating characteristic curve was used for comparison across and within model classes. All analyses were performed in MATLAB using a MacBook Pro.RESULTSThe authors included 6 preoperative imaging and demographic variables: tumor volume, degree of peritumoral edema, presence of necrosis, tumor location, patient sex, and presence of a draining vein to construct the models. The artificial neural networks outperformed all other ML models across the true-positive versus false-positive (receiver operating characteristic) space (area under curve = 0.8895).CONCLUSIONSML algorithms are powerful computational tools that can predict meningioma grade with great accuracy.

Using artificial neural networks to identify patients with concussion and postconcussion syndrome based on antisaccades

OBJECTIVE

Artificial neural networks (ANNs) have shown considerable promise as decision support tools in medicine, including neurosurgery. However, their use in concussion and postconcussion syndrome (PCS) has been limited. The authors explore the value of using an ANN to identify patients with concussion/PCS based on their antisaccade performance.

METHODS

Study participants were prospectively recruited from the emergency department and head injury clinic of a large teaching hospital in Toronto. Acquaintances of study participants were used as controls. Saccades were measured using an automated, portable, head-mounted device preprogrammed with an antisaccade task. Each participant underwent 100 trials of the task and 11 saccade parameters were recorded for each trial. ANN analysis was performed using the MATLAB Neural Network Toolbox, and individual saccade parameters were further explored with receiver operating characteristic (ROC) curves and a logistic regression analysis.

RESULTS

Control (n = 15), concussion (n = 32), and PCS (n = 25) groups were matched by age and level of education. The authors examined 11 saccade parameters and found that the prosaccade error rate (p = 0.04) and median antisaccade latency (p = 0.02) were significantly different between control and concussion/PCS groups. When used to distinguish concussion and PCS participants from controls, the neural networks achieved accuracies of 67% and 72%, respectively. This method was unable to distinguish study patients with concussion from those with PCS, suggesting persistence of eye movement abnormalities in patients with PCS. The authors' observations also suggest the potential for improved results with a larger training sample.

CONCLUSIONS

This study explored the utility of ANNs in the diagnosis of concussion/PCS based on antisaccades. With the use of an ANN, modest accuracy was achieved in a small cohort. In addition, the authors explored the pearls and pitfalls of this novel approach and identified important future directions for this research.

Machine Learning to Predict In-Hospital Morbidity and Mortality after Traumatic Brain Injury

Recently, successful predictions using machine learning (ML) algorithms have been reported in various fields. However, in traumatic brain injury (TBI) cohorts, few studies have examined modern ML algorithms. To develop a simple ML model for TBI outcome prediction, we conducted a performance comparison of nine algorithms: ridge regression, least absolute shrinkage and selection operator (LASSO) regression, random forest, gradient boosting, extra trees, decision tree, Gaussian naïve Bayes, multi-nomial naïve Bayes, and support vector machine. Fourteen feasible parameters were introduced in the ML models, including age, Glasgow Coma Scale (GCS), systolic blood pressure (SBP), abnormal pupillary response, major extracranial injury, computed tomography (CT) findings, and routinely collected laboratory values (glucose, C-reactive protein [CRP], and fibrin/fibrinogen degradation products [FDP]). Data from 232 patients with TBI were randomly divided into a training sample (80%) for hyperparameter tuning and validation sample (20%). The bootstrap method was used for validation. Random forest demonstrated the best performance for in-hospital poor outcome prediction and ridge regression for in-hospital mortality prediction: the mean statistical measures were 100% sensitivity, 72.3% specificity, 91.7% accuracy, and 0.895 area under the receiver operating characteristic curve (AUC); and 88.4% sensitivity, 88.2% specificity, 88.6% accuracy, and 0.875 AUC, respectively. Based on the feature selection method using the tree-based ensemble algorithm, age, Glasgow Coma Scale, fibrin/fibrinogen degradation products, and glucose were identified as the most important prognostic factors for poor outcome and mortality. Our results indicate the relatively good predictive performance of modern ML for TBI outcome. Further external validation is required for more heterogeneous samples to confirm our results.

Natural and Artificial Intelligence in Neurosurgery: A Systematic Review

BACKGROUND

Machine learning (ML) is a domain of artificial intelligence that allows computer algorithms to learn from experience without being explicitly programmed.

OBJECTIVE

To summarize neurosurgical applications of ML where it has been compared to clinical expertise, here referred to as "natural intelligence."

METHODS

A systematic search was performed in the PubMed and Embase databases as of August 2016 to review all studies comparing the performance of various ML approaches with that of clinical experts in neurosurgical literature.

RESULTS

Twenty-three studies were identified that used ML algorithms for diagnosis, presurgical planning, or outcome prediction in neurosurgical patients. Compared to clinical experts, ML models demonstrated a median absolute improvement in accuracy and area under the receiver operating curve of 13% (interquartile range 4-21%) and 0.14 (interquartile range 0.07-0.21), respectively. In 29 (58%) of the 50 outcome measures for which a P-value was provided or calculated, ML models outperformed clinical experts (P < .05). In 18 of 50 (36%), no difference was seen between ML and expert performance (P > .05), while in 3 of 50 (6%) clinical experts outperformed ML models (P < .05). All 4 studies that compared clinicians assisted by ML models vs clinicians alone demonstrated a better performance in the first group.

CONCLUSION

We conclude that ML models have the potential to augment the decision-making capacity of clinicians in neurosurgical applications; however, significant hurdles remain associated with creating, validating, and deploying ML models in the clinical setting. Shifting from the preconceptions of a human-vs-machine to a human-and-machine paradigm could be essential to overcome these hurdles.

Machine learning applications to clinical decision support in neurosurgery: an artificial intelligence augmented systematic review

Machine learning (ML) involves algorithms learning patterns in large, complex datasets to predict and classify. Algorithms include neural networks (NN), logistic regression (LR), and support vector machines (SVM). ML may generate substantial improvements in neurosurgery. This systematic review assessed the current state of neurosurgical ML applications and the performance of algorithms applied. Our systematic search strategy yielded 6866 results, 70 of which met inclusion criteria. Performance statistics analyzed included area under the receiver operating characteristics curve (AUC), accuracy, sensitivity, and specificity. Natural language processing (NLP) was used to model topics across the corpus and to identify keywords within surgical subspecialties. ML applications were heterogeneous. The densest cluster of studies focused on preoperative evaluation, planning, and outcome prediction in spine surgery. The main algorithms applied were NN, LR, and SVM. Input and output features varied widely and were listed to facilitate future research. The accuracy (F_(2,19) = 6.56, p < 0.01) and specificity (F_(2,16) = 5.57, p < 0.01) of NN, LR, and SVM differed significantly. NN algorithms demonstrated significantly higher accuracy than LR. SVM demonstrated significantly higher specificity than LR. We found no significant difference between NN, LR, and SVM AUC and sensitivity. NLP topic modeling reached maximum coherence at seven topics, which were defined by modeling approach, surgery type, and pathology themes. Keywords captured research foci within surgical domains. ML technology accurately predicts outcomes and facilitates clinical decision-making in neurosurgery. NNs frequently outperformed other algorithms on supervised learning tasks. This study identified gaps in the literature and opportunities for future neurosurgical ML research.

Machine learning applications for the prediction of surgical site infection in neurological operations

OBJECTIVE

Surgical site infection (SSI) following a neurosurgical operation is a complication that impacts morbidity, mortality, and economics. Currently, machine learning (ML) algorithms are used for outcome prediction in various neurosurgical aspects. The implementation of ML algorithms to learn from medical data may help in obtaining prognostic information on diseases, especially SSIs. The purpose of this study was to compare the performance of various ML models for predicting surgical infection after neurosurgical operations.

METHODS

A retrospective cohort study was conducted on patients who had undergone neurosurgical operations at tertiary care hospitals between 2010 and 2017. Supervised ML algorithms, which included decision tree, naive Bayes with Laplace correction, k-nearest neighbors, and artificial neural networks, were trained and tested as binary classifiers (infection or no infection). To evaluate the ML models from the testing data set, their sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), as well as their accuracy, receiver operating characteristic curve, and area under the receiver operating characteristic curve (AUC) were analyzed.

RESULTS

Data were available for 1471 patients in the study period. The SSI rate was 4.6%, and the type of SSI was superficial, deep, and organ/space in 1.2%, 0.8%, and 2.6% of cases, respectively. Using the backward stepwise method, the authors determined that the significant predictors of SSI in the multivariable Cox regression analysis were postoperative CSF leakage/subgaleal collection (HR 4.24, p < 0.001) and postoperative fever (HR 1.67, p = 0.04). Compared with other ML algorithms, the naive Bayes had the highest performance with sensitivity at 63%, specificity at 87%, PPV at 29%, NPV at 96%, and AUC at 76%.

CONCLUSIONS

The naive Bayes algorithm is highlighted as an accurate ML method for predicting SSI after neurosurgical operations because of its reasonable accuracy. Thus, it can be used to effectively predict SSI in individual neurosurgical patients. Therefore, close monitoring and allocation of treatment strategies can be informed by ML predictions in general practice.

Pharmacokinetics and efficacy of PT302, a sustained-release Exenatide formulation, in a murine model of mild traumatic brain injury

Traumatic brain injury (TBI) is a neurodegenerative disorder for which no effective pharmacological treatment is available. Glucagon-like peptide 1 (GLP-1) analogues such as Exenatide have previously demonstrated neurotrophic and neuroprotective effects in cellular and animal models of TBI. However, chronic or repeated administration was needed for efficacy. In this study, the pharmacokinetics and efficacy of PT302, a clinically available sustained-release Exenatide formulation (SR-Exenatide) were evaluated in a concussive mild (m)TBI mouse model. A single subcutaneous (s.c.) injection of PT302 (0.6, 0.12, and 0.024 mg/kg) was administered and plasma Exenatide concentrations were time-dependently measured over 3 weeks. An initial rapid regulated release of Exenatide in plasma was followed by a secondary phase of sustained-release in a dose-dependent manner. Short- and longer-term (7 and 30 day) cognitive impairments (visual and spatial deficits) induced by weight drop mTBI were mitigated by a single post-injury treatment with Exenatide delivered by s.c. injection of PT302 in clinically translatable doses. Immunohistochemical evaluation of neuronal cell death and inflammatory markers, likewise, cross-validated the neurotrophic and neuroprotective effects of SR-Exenatide in this mouse mTBI model. Exenatide central nervous system concentrations were 1.5% to 2.0% of concomitant plasma levels under steady-state conditions. These data demonstrate a positive beneficial action of PT302 in mTBI. This convenient single, sustained-release dosing regimen also has application for other neurological disorders, such as Alzheimer's disease, Parkinson's disease, multiple system atrophy and multiple sclerosis where prior preclinical studies, likewise, have demonstrated positive Exenatide actions.

Early Identification of High-Risk TIA or Minor Stroke Using Artificial Neural Network

Background and Purpose: The risk of recurrent stroke following a transient ischemic attack (TIA) or minor stroke is high, despite of a significant reduction in the past decade. In this study, we investigated the feasibility of using artificial neural network (ANN) for risk stratification of TIA or minor stroke patients.

Methods: Consecutive patients with acute TIA or minor ischemic stroke presenting at a tertiary hospital during a 2-year period were recruited. We collected demographics, clinical and imaging data at baseline. The primary outcome was recurrent ischemic stroke within 1 year. We developed ANN models to predict the primary outcome. We randomly down-sampled patients without a primary outcome to 1:1 match with those with a primary outcome to mitigate data imbalance. We used a 5-fold cross-validation approach to train and test the ANN models to avoid overfitting. We employed 19 independent variables at baseline as the input neurons in the ANN models, using a learning algorithm based on backpropagation to minimize the loss function. We obtained the sensitivity, specificity, accuracy and the c statistic of each ANN model from the 5 rounds of cross-validation and compared that of support vector machine (SVM) and Naïve Bayes classifier in risk stratification of the patients.

Results: A total of 451 acute TIA or minor stroke patients were enrolled. Forty (8.9%) patients had a recurrent ischemic stroke within 1 year. Another 40 patients were randomly selected from those with no recurrent stroke, so that data from 80 patients in total were used for 5 rounds of training and testing of ANN models. The median sensitivity, specificity, accuracy and c statistic of the ANN models to predict recurrent stroke at 1 year was 75%, 75%, 75%, and 0.77, respectively. ANN model outperformed SVM and Naïve Bayes classifier in our dataset for predicting relapse after TIA or minor stroke.

Conclusion: This pilot study indicated that ANN may yield a novel and effective method in risk stratification of TIA and minor stroke. Further studies are warranted for verification and improvement of the current ANN model.

Comparison of artificial neural network and logistic regression models for prediction of outcomes in trauma patients: A systematic review and meta-analysis

BACKGROUND

Currently, two models of artificial neural network (ANN) and logistic regression (LR) are known as models that extensively used in medical sciences. The aim of this study was to compare the ANN and LR models in prediction of Health-related outcomes in traumatic patients using a systematic review.

METHODS

The study was planned and conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) checklist. A literature search of published studies was conducted using PubMed, Embase, Web of knowledge, Scopus, and Google Scholar in May 2018. Joanna Briggs Institute (JBI) checklists was used for assessing the quality of the included articles.

RESULTS

The literature searches yielded 326 potentially relevant studies from the primary searches. Overall, the review included 10 unique studies. The results of this study showed that the area under curve (AUC) for the ANN was 0.91, (95% CI 0.89-0.83) and 0.89, (95% CI 0.87-90) for the LR in random effect model. The accuracy rate for ANN and LR in random effect models were 90.5, (95% CI, 87.6-94.2) and 83.2, (95% CI 75.1-91.2), respectively.

CONCLUSION

The results of our study showed that ANN has better performance than LR in predicting the terminal outcomes of traumatic patients in both the AUC and accuracy rate. Using an ANN to predict the final implications of trauma patients can provide more accurate clinical decisions.

A performance analysis of Brazilian public health: TOPSIS and neural networks application

Purpose

The purpose of this paper is to estimate the performance of Brazilian hospitals’ services and to examine contextual variables in the socioeconomic, demographic and institutional domains as predictors of the performance levels attained.

Design/methodology/approach

The paper applied a two-stage approach of the technique for order preference by similarity to the ideal solution (TOPSIS) in public hospitals in 92 Rio de Janeiro municipalities, covering the 2008–2013 period. First, TOPSIS is used to estimate the relative performance of hospitals in each municipality. Next, TOPSIS results are combined with neural networks in an effort to originate a performance model with predictive ability. Data refer to hospitals’ outpatient and inpatient services, based on frequent indicators adopted by the healthcare literature.

Findings

Despite a slight performance increase over the period, substantial room for improvement is observed. The most important performance predictors were related to the demographic and socioeconomic status (area in square feet and GDP per capita) and to the juridical nature and type of ownership of the healthcare facilities (number of federal and private hospitals).

Practical implications

The results provide managerial insights regarding the performance of public hospitals and opportunities for better resource allocation in the healthcare sector. The paper also considers the impact of external socioeconomic, demographic and institutional factors on hospitals’ performance, indicating the importance of integrative public health policies.

Originality/value

This study displays an innovative context for applying the two-stage TOPSIS technique, with similar efforts not having been identified in the healthcare literature.

Neuronal Enriched Extracellular Vesicle Proteins as Biomarkers for Traumatic Brain Injury

Traumatic brain injury (TBI) is a major cause of injury-related death throughout the world and lacks effective treatment. Surviving TBI patients often develop neuropsychiatric symptoms, and the molecular mechanisms underlying the neuronal damage and recovery following TBI are not well understood. Extracellular vesicles (EVs) are membranous nanoparticles that are divided into exosomes (originating in the endosomal/multi-vesicular body [MVB] system) and microvesicles (larger EVs produced through budding of the plasma membrane). Both types of EVs are generated by all cells and are secreted into the extracellular environment, and participate in cell-to-cell communication and protein and RNA delivery. EVs enriched for neuronal origin can be harvested from peripheral blood samples and their contents quantitatively examined as a window to follow potential changes occurring in brain. Recent studies suggest that the levels of exosomal proteins and microRNAs (miRNAs) may represent novel biomarkers to support the clinical diagnosis and potential response to treatment for neurological disorders. In this review, we focus on the biogenesis of EVs, their molecular composition, and recent advances in research of their contents as potential diagnostic tools for TBI.

Real and predicted mortality under health spending constraints in Italy: a time trend analysis through artificial neural networks

BACKGROUND

After 2008 global economic crisis, Italian governments progressively reduced public healthcare financing. Describing the time trend of health outcomes and health expenditure may be helpful for policy makers during the resources' allocation decision making process. The aim of this paper is to analyze the trend of mortality and health spending in Italy and to investigate their correlation in consideration of the funding constraints experienced by the Italian national health system (SSN).

METHODS

We conducted a 20-year time-series study. Secondary data has been extracted from a national, institution based and publicly accessible retrospective database periodically released by the Italian Institute of Statistics. Age standardized all-cause mortality rate (MR) and health spending (Directly Provided Services - DPS, Agreed-Upon Services - TAUS, and private expenditure) were reviewed. Time trend analysis (1995-2014) through OLS and Multilayer Feed-forward Neural Networks (MFNN) models to forecast mortality and spending trend was performed. The association between healthcare expenditure and MR was analyzed through a fixed effect regression model. We then repeated MFNN time trend forecasting analyses on mortality by adding the spending item resulted significantly related with MR in the fixed effect analyses.

RESULTS

DPS and TAUS decreased since 2011. There was a mismatch in mortality rates between real and predicted values. DPS resulted significantly associated to mortality (p < 0.05). In repeated mortality forecasting analysis, predicted MR was found to be lower when considering the pre-constraints health spending trend.

CONCLUSIONS

Between 2011 and 2014, Italian public health spending items showed a reduction when compared to prior years. Spending on services directly provided free of charge appears to be the financial driving force of the Italian public health system. The overall mortality was found to be higher than the predicted trend and this scenario may be partially attributable to the healthcare funding constraints experienced by the SSN.

Machine Learning for Predicting Outcomes in Trauma

To date, there are no reviews on machine learning (ML) for predicting outcomes in trauma. Consequently, it remains unclear as to how ML-based prediction models compare in the triage and assessment of trauma patients. The objective of this review was to survey and identify studies involving ML for predicting outcomes in trauma, with the hypothesis that models predicting similar outcomes may share common features but the performance of ML in these studies will differ greatly. MEDLINE and other databases were searched for studies involving trauma and ML. Sixty-five observational studies involving ML for the prediction of trauma outcomes met inclusion criteria. In total 2,433,180 patients were included in the studies. The studies focused on prediction of the following outcome measures: survival/mortality (n = 34), morbidity/shock/hemorrhage (n = 12), hospital length of stay (n = 7), hospital admission/triage (n = 6), traumatic brain injury (n = 4), life-saving interventions (n = 5), post-traumatic stress disorder (n = 4), and transfusion (n = 1). Six studies were prospective observational studies. Of the 65 studies, 33 used artificial neural networks for prediction. Importantly, most studies demonstrated the benefits of ML models. However, algorithm performance was assessed differently by different authors. Sensitivity-specificity gap values varied greatly from 0.035 to 0.927. Notably, studies shared many features for model development. A common ML feature base may be determined for predicting outcomes in trauma. However, the impact of ML will require further validation in prospective observational studies and randomized clinical trials, establishment of common performance criteria, and high-quality evidence about clinical and economic impacts before ML can be widely accepted in practice.

Impact of grouping complications on mortality in traumatic brain injury: A nationwide population-based study

Traumatic brain injury (TBI) is an important health issue with high mortality. Various complications of physiological and cognitive impairment may result in disability or death after TBI. Grouping of these complications could be treated as integrated post-TBI syndromes. To improve risk estimation, grouping TBI complications should be investigated, to better predict TBI mortality. This study aimed to estimate mortality risk based on grouping of complications among TBI patients. Taiwan's National Health Insurance Research Database was used in this study. TBI was defined according to the International Classification of Diseases, Ninth Revision, Clinical Modification codes: 801-804 and 850-854. The association rule data mining method was used to analyze coexisting complications after TBI. The mortality risk of post-TBI complication sets with the potential risk factors was estimated using Cox regression. A total 139,254 TBI patients were enrolled in this study. Intracerebral hemorrhage was the most common complication among TBI patients. After frequent item set mining, the most common post-TBI grouping of complications comprised pneumonia caused by acute respiratory failure (ARF) and urinary tract infection, with mortality risk 1.55 (95% C.I.: 1.51-1.60), compared with those without the selected combinations. TBI patients with the combined combinations have high mortality risk, especially those aged <20 years with septicemia, pneumonia, and ARF (HR: 4.95, 95% C.I.: 3.55-6.88). We used post-TBI complication sets to estimate mortality risk among TBI patients. According to the combinations determined by mining, especially the combination of septicemia with pneumonia and ARF, TBI patients have a 1.73-fold increased mortality risk, after controlling for potential demographic and clinical confounders. TBI patients aged<20 years with each combination of complications also have increased mortality risk. These results could provide physicians and caregivers with important information to increase their awareness about sequences of clinical syndromes among TBI patients, to prevent possible deaths among these patients.

An introduction and overview of machine learning in neurosurgical care

BACKGROUND

Machine learning (ML) is a branch of artificial intelligence that allows computers to learn from large complex datasets without being explicitly programmed. Although ML is already widely manifest in our daily lives in various forms, the considerable potential of ML has yet to find its way into mainstream medical research and day-to-day clinical care. The complex diagnostic and therapeutic modalities used in neurosurgery provide a vast amount of data that is ideally suited for ML models. This systematic review explores ML's potential to assist and improve neurosurgical care.

METHOD

A systematic literature search was performed in the PubMed and Embase databases to identify all potentially relevant studies up to January 1, 2017. All studies were included that evaluated ML models assisting neurosurgical treatment.

RESULTS

Of the 6,402 citations identified, 221 studies were selected after subsequent title/abstract and full-text screening. In these studies, ML was used to assist surgical treatment of patients with epilepsy, brain tumors, spinal lesions, neurovascular pathology, Parkinson's disease, traumatic brain injury, and hydrocephalus. Across multiple paradigms, ML was found to be a valuable tool for presurgical planning, intraoperative guidance, neurophysiological monitoring, and neurosurgical outcome prediction.

CONCLUSIONS

ML has started to find applications aimed at improving neurosurgical care by increasing the efficiency and precision of perioperative decision-making. A thorough validation of specific ML models is essential before implementation in clinical neurosurgical care. To bridge the gap between research and clinical care, practical and ethical issues should be considered parallel to the development of these techniques.

Artificial neural network predicts hemorrhagic contusions following decompressive craniotomy in traumatic brain injury

BACKGROUND

This study aimed to explore relevant factors of hemorrhagic contusions following decompressive craniotomy (DC) in traumatic brain injury (TBI) and create an artificial neural network (ANN) prediction model of the risk factors of hemorrhagic contusions.

METHODS

This study analyzed 425 patients with TBI who underwent DC in the Neurosurgery Department of Shaoxing People's Hospital between 2009 and 2014. Patients were divided into two groups according to the first postoperative CT scans: hemorrhage group and non-hemorrhage group. Gender, age, preoperative situations (Initial Rotterdam CT Score, GCS Score, pupillary response, laboratory data and preoperative hematoma), the time gap between trauma and DC, postoperative CT scans, and Glasgow Outcome Scale (GOS) scores were recorded. ANN was used to predict hematoma. Correlation analysis was used to state the relationship between increased hemorrhage volumes and GOS scores.

RESULTS

The ANN prediction model was established. This model included 11 parameters: initial Rotterdam CT score, GCS score, C-reactive protein, age, the time gap between trauma and DC, pupillary response, platelet count, bone-flap size, glucose level, hernia magnitude and preoperative hematoma volume. The overall predictive accuracy of the model was 73.0%.

CONCLUSIONS

Initial Rotterdam CT scores and GCS scores may predict the risk of expansion contusions following DC. The ANN prediction model has a high accuracy to forecast hemorrhage.

The impact of presurgical comorbidities on discharge disposition and length of hospitalization following craniotomy for brain tumor

BACKGROUND

Identifying risk factors for negative postoperative outcomes is an important part of providing quality care. Here, we build machine learning (ML) ensembles to model the independent impact of presurgical comorbidities on discharge disposition and length of stay (LOS) following brain tumor resection from the HCUP National Inpatient Sample (NIS).

METHODS

We performed a retrospective cohort study of 41,222 patients who underwent craniotomy for brain tumors during 2002-2011 and were registered in the NIS. Twenty-six ML algorithms were trained on prehospitalization variables to predict nonhome discharge and extended LOS (>7 days), and the most predictive algorithms combined to create ensemble models. Models were validated to demonstrate generalizability. Analysis was done to identify which and how specific comorbidities influence ensemble predictions.

RESULTS

Receiver operating curve analysis showed area under the curve of 0.796 and 0.824 for the disposition and LOS ensembles, respectively. The disposition ensemble was most strongly influenced by preoperative paralysis and fluid/electrolyte abnormalities, which independently increased the risk of nonhome discharge in craniotomy patients by 35.4% and 13.9%, respectively. The LOS ensemble was most strongly influenced by the presence of preoperative paralysis, fluid/electrolyte abnormalities, and other nonparalysis neurological deficits, which independently increased the risk of extended LOS in craniotomy patients by 20.4%, 22.5%, and 38.3%, respectively.

CONCLUSIONS

In this study, we used ML ensembles to identify preoperative comorbidities that increased the risk of nonhome discharge and extended LOS following craniotomy for brain tumor. Recognizing these risk factors for poor postsurgical outcomes can improve patient counseling and offer opportunities for quality improvement.

Using a Guided Machine Learning Ensemble Model to Predict Discharge Disposition following Meningioma Resection

Objective  Machine learning (ML) algorithms are powerful tools for predicting patient outcomes. This study pilots a novel approach to algorithm selection and model creation using prediction of discharge disposition following meningioma resection as a proof of concept. Materials and Methods  A diversity of ML algorithms were trained on a single-institution database of meningioma patients to predict discharge disposition. Algorithms were ranked by predictive power and top performers were combined to create an ensemble model. The final ensemble was internally validated on never-before-seen data to demonstrate generalizability. The predictive power of the ensemble was compared with a logistic regression. Further analyses were performed to identify how important variables impact the ensemble. Results  Our ensemble model predicted disposition significantly better than a logistic regression (area under the curve of 0.78 and 0.71, respectively, p  = 0.01). Tumor size, presentation at the emergency department, body mass index, convexity location, and preoperative motor deficit most strongly influence the model, though the independent impact of individual variables is nuanced. Conclusion  Using a novel ML technique, we built a guided ML ensemble model that predicts discharge destination following meningioma resection with greater predictive power than a logistic regression, and that provides greater clinical insight than a univariate analysis. These techniques can be extended to predict many other patient outcomes of interest.

The Impact of Race on Discharge Disposition and Length of Hospitalization After Craniotomy for Brain Tumor

BACKGROUND

Racial disparities exist in health care, frequently resulting in unfavorable outcomes for minority patients. Here, we use guided machine learning (ML) ensembles to model the impact of race on discharge disposition and length of stay (LOS) after brain tumor surgery from the Healthcare Cost and Utilization Project National Inpatient Sample.

METHODS

We performed a retrospective cohort study of 41,222 patients who underwent craniotomies for brain tumors from 2002 to 2011 and were registered in the National Inpatient Sample. Twenty-six ML algorithms were trained on prehospitalization variables to predict non-home discharge and extended LOS (>7 days) after brain tumor resection, and the most predictive algorithms combined to create ensemble models. Partial dependence analysis was performed to measure the independent impact of race on the ensembles.

RESULTS

The guided ML ensembles predicted non-home disposition (area under the curve, 0.796) and extended LOS (area under the curve, 0.824) with good discrimination. Partial dependence analysis showed that black race increases the risk of non-home discharge and extended LOS over white race by 6.9% and 6.5%, respectively. Other, nonblack race increases the risk of extended LOS over white race by 6.0%. The impact of race on these outcomes is not seen when analyzing the general inpatient or general operative population.

CONCLUSIONS

Minority race independently increases the risk of extended LOS and black race increases the risk of non-home discharge in patients undergoing brain tumor resection, a finding not mimicked in the general inpatient or operative population. Recognition of the influence of race on discharge and LOS could generate interventions that may improve outcomes in this population.