Application of an artificial neural network to predict postinduction hypotension during general anesthesia

Predictive ability of hypotension prediction index and machine learning methods in intraoperative hypotension: a systematic review and meta-analysis

INTRODUCTION

Intraoperative Hypotension (IOH) poses a substantial risk during surgical procedures. The integration of Artificial Intelligence (AI) in predicting IOH holds promise for enhancing detection capabilities, providing an opportunity to improve patient outcomes. This systematic review and meta analysis explores the intersection of AI and IOH prediction, addressing the crucial need for effective monitoring in surgical settings.

METHOD

A search of Pubmed, Scopus, Web of Science, and Embase was conducted. Screening involved two-phase assessments by independent reviewers, ensuring adherence to predefined PICOS criteria. Included studies focused on AI models predicting IOH in any type of surgery. Due to the high number of studies evaluating the hypotension prediction index (HPI), we conducted two sets of meta-analyses: one involving the HPI studies and one including non-HPI studies. In the HPI studies the following outcomes were analyzed: cumulative duration of IOH per patient, time weighted average of mean arterial pressure < 65 (TWA-MAP < 65), area under the threshold of mean arterial pressure (AUT-MAP), and area under the receiver operating characteristics curve (AUROC). In the non-HPI studies, we examined the pooled AUROC of all AI models other than HPI.

RESULTS

43 studies were included in this review. Studies showed significant reduction in IOH duration, TWA-MAP < 65 mmHg, and AUT-MAP < 65 mmHg in groups where HPI was used. AUROC for HPI algorithms demonstrated strong predictive performance (AUROC = 0.89, 95CI). Non-HPI models had a pooled AUROC of 0.79 (95CI: 0.74, 0.83).

CONCLUSION

HPI demonstrated excellent ability to predict hypotensive episodes and hence reduce the duration of hypotension. Other AI models, particularly those based on deep learning methods, also indicated a great ability to predict IOH, while their capacity to reduce IOH-related indices such as duration remains unclear.

Artificial intelligence and its clinical application in Anesthesiology: a systematic review

PURPOSE

Application of artificial intelligence (AI) in medicine is quickly expanding. Despite the amount of evidence and promising results, a thorough overview of the current state of AI in clinical practice of anesthesiology is needed. Therefore, our study aims to systematically review the application of AI in this context.

METHODS

A systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We searched Medline and Web of Science for articles published up to November 2022 using terms related with AI and clinical practice of anesthesiology. Articles that involved animals, editorials, reviews and sample size lower than 10 patients were excluded. Characteristics and accuracy measures from each study were extracted.

RESULTS

A total of 46 articles were included in this review. We have grouped them into 4 categories with regard to their clinical applicability: (1) Depth of Anesthesia Monitoring; (2) Image-guided techniques related to Anesthesia; (3) Prediction of events/risks related to Anesthesia; (4) Drug administration control. Each group was analyzed, and the main findings were summarized. Across all fields, the majority of AI methods tested showed superior performance results compared to traditional methods.

CONCLUSION

AI systems are being integrated into anesthesiology clinical practice, enhancing medical professionals' skills of decision-making, diagnostic accuracy, and therapeutic response.

Muscle-related parameters-based machine learning model for predicting postinduction hypotension in patients undergoing colorectal tumor resection surgery

Objectives

This study used machine learning algorithms to identify important variables and predict postinduction hypotension (PIH) in patients undergoing colorectal tumor resection surgery.

Methods

Data from 318 patients who underwent colorectal tumor resection under general anesthesia were analyzed. The training and test sets are divided based on the timeline. The Boruta algorithm was used to screen relevant basic characteristic variables and establish a model for the training set. Four models, regression tree, K-nearest neighbor, neural network, and random forest (RF), were built using repeated cross-validation and hyperparameter optimization. The best model was selected, and a sorting chart of the feature variables, a univariate partial dependency profile, and a breakdown profile were drawn. R2, mean absolute error (MAE), mean squared error (MSE), and root MSE (RMSE) were used to plot regression fitting curves for the training and test sets.

Results

The basic feature variables associated with the Boruta screening were age, sex, body mass index, L3 skeletal muscle index, and HUAC. In the optimal RF model, R2 was 0.7708 and 0.7591, MAE was 0.0483 and 0.0408, MSE was 0.0038 and 0.0028, and RMSE was 0.0623 and 0.0534 for the training and test sets, respectively.

Conclusion

A high-performance algorithm was established and validated to demonstrate the degree of change in blood pressure after induction to control important characteristic variables and reduce PIH occurrence.

Effective hemodynamic monitoring

Hemodynamic monitoring is the centerpiece of patient monitoring in acute care settings. Its effectiveness in terms of improved patient outcomes is difficult to quantify. This review focused on effectiveness of monitoring-linked resuscitation strategies from: (1) process-specific monitoring that allows for non-specific prevention of new onset cardiovascular insufficiency (CVI) in perioperative care. Such goal-directed therapy is associated with decreased perioperative complications and length of stay in high-risk surgery patients. (2) Patient-specific personalized resuscitation approaches for CVI. These approaches including dynamic measures to define volume responsiveness and vasomotor tone, limiting less fluid administration and vasopressor duration, reduced length of care. (3) Hemodynamic monitoring to predict future CVI using machine learning approaches. These approaches presently focus on predicting hypotension. Future clinical trials assessing hemodynamic monitoring need to focus on process-specific monitoring based on modifying therapeutic interventions known to improve patient-centered outcomes.

Predicting intraoperative hypotension using deep learning with waveforms of arterial blood pressure, electroencephalogram, and electrocardiogram: Retrospective study

To develop deep learning models for predicting Interoperative hypotension (IOH) using waveforms from arterial blood pressure (ABP), electrocardiogram (ECG), and electroencephalogram (EEG), and to determine whether combination ABP with EEG or CG improves model performance. Data were retrieved from VitalDB, a public data repository of vital signs taken during surgeries in 10 operating rooms at Seoul National University Hospital from January 6, 2005, to March 1, 2014. Retrospective data from 14,140 adult patients undergoing non-cardiac surgery with general anaesthesia were used. The predictive performances of models trained with different combinations of waveforms were evaluated and compared at time points at 3, 5, 10, 15 minutes before the event. The performance was calculated by area under the receiver operating characteristic (AUROC), area under the precision-recall curve (AUPRC), sensitivity and specificity. The model performance was better in the model using both ABP and EEG waveforms than in all other models at all time points (3, 5, 10, and 15 minutes before an event) Using high-fidelity ABP and EEG waveforms, the model predicted IOH with a AUROC and AUPRC of 0.935 [0.932 to 0.938] and 0.882 [0.876 to 0.887] at 5 minutes before an IOH event. The output of both ABP and EEG was more calibrated than that using other combinations or ABP alone. The results demonstrate that a predictive deep neural network can be trained using ABP, ECG, and EEG waveforms, and the combination of ABP and EEG improves model performance and calibration.

Machine Learning, Deep Learning, and Closed Loop Devices-Anesthesia Delivery

With the tremendous volume of data captured during surgeries and procedures, critical care, and pain management, the field of anesthesiology is uniquely suited for the application of machine learning, neural networks, and closed loop technologies. In the past several years, this area has expanded immensely in both interest and clinical applications. This article provides an overview of the basic tenets of machine learning, neural networks, and closed loop devices, with emphasis on the clinical applications of these technologies.

One of the first validations of an artificial intelligence algorithm for clinical use: The impact on intraoperative hypotension prediction and clinical decision-making

This review describes the steps and conclusions from the development and validation of an artificial intelligence algorithm (the Hypotension Prediction Index), one of the first machine learning predictive algorithms used in the operating room environment. The algorithm has been demonstrated to reduce intraoperative hypotension in two randomized controlled trials via real-time prediction of upcoming hypotensive events prompting anesthesiologists to act earlier, more often, and differently in managing impending hypotension. However, the algorithm entails no dynamic learning process that evolves from use in clinical patient care, meaning the algorithm is fixed, and furthermore provides no insight into the decisional process that leads to an early warning for intraoperative hypotension, which makes the algorithm a "black box." Many other artificial intelligence machine learning algorithms have these same disadvantages. Clinical validation of such algorithms is relatively new and requires more standardization, as guidelines are lacking or only now start to be drafted. Before adaptation in clinical practice, impact of artificial intelligence algorithms on clinical behavior, outcomes and economic advantages should be studied too.

Artificial Intelligence in Anesthesiology: Current Techniques, Clinical Applications, and Limitations

Artificial intelligence has been advancing in fields including anesthesiology. This scoping review of the intersection of artificial intelligence and anesthesia research identified and summarized six themes of applications of artificial intelligence in anesthesiology: (1) depth of anesthesia monitoring, (2) control of anesthesia, (3) event and risk prediction, (4) ultrasound guidance, (5) pain management, and (6) operating room logistics. Based on papers identified in the review, several topics within artificial intelligence were described and summarized: (1) machine learning (including supervised, unsupervised, and reinforcement learning), (2) techniques in artificial intelligence (e.g., classical machine learning, neural networks and deep learning, Bayesian methods), and (3) major applied fields in artificial intelligence.The implications of artificial intelligence for the practicing anesthesiologist are discussed as are its limitations and the role of clinicians in further developing artificial intelligence for use in clinical care. Artificial intelligence has the potential to impact the practice of anesthesiology in aspects ranging from perioperative support to critical care delivery to outpatient pain management.

Development of artificial neural network model for prediction of post-streptococcus mutans in dental caries

BACKGROUND AND OBJECTIVE

Streptococcus mutans is the primary initiator and most common organism associated with dental caries. Prediction of post-Streptococcus mutans favours in the selection of appropriate caries excavation method which eventually results in meliorate caries-free cavity preparation for restoration. The objective of this study is to predict the post-Streptococcus mutans prior to dental caries excavation based on pre- Streptococcus mutans using iOS App developed on Artificial Neural Network (ANN) model.

METHODS

For the current research work, children with occlusal dentinal caries lesion were chosen, 45 primary molar teeth cases were studied. Caries excavation was done with carbide bur, polymer bur and spoon excavator. The colony forming units for pre and post-Streptococcus mutans were recorded, data emanating from clinical trials was employed to develop the ANN models. ANN models were trained, validated and tested with the registered clinical data using different ANN architectures.

RESULTS

Feedforward backpropagation ANN model with an architecture of 4-5-1, predicts post-Streptococcus mutans with an efficiency of 0.99033, mean squared error and mean absolute percentage error for testing cases were 0.2341 and 4.967 respectively.

CONCLUSIONS

Caries excavation methods and pre-Streptococcus mutans are feed as inputs, while post-Streptococcus mutans as targets to develop ANN model. Based on the developed ANN model, an ingenious iOS App was developed, the global clinician may utilize the App to meticulously predict post-Streptococcus mutans on iPhone based on pre-Streptococcus mutans, which in turn aids in decision making for the selection of caries excavation method. This study manifests the potential application of iOS App with built-in ANN model in efficiently predicting the post-Streptococcus mutans. Also, the study extends scope for applications of iOS App with built-in ANN models in clinical medicine.

Artificial Intelligence and Machine Learning in Anesthesiology

Commercial applications of artificial intelligence and machine learning have made remarkable progress recently, particularly in areas such as image recognition, natural speech processing, language translation, textual analysis, and self-learning. Progress had historically languished in these areas, such that these skills had come to seem ineffably bound to intelligence. However, these commercial advances have performed best at single-task applications in which imperfect outputs and occasional frank errors can be tolerated.The practice of anesthesiology is different. It embodies a requirement for high reliability, and a pressured cycle of interpretation, physical action, and response rather than any single cognitive act. This review covers the basics of what is meant by artificial intelligence and machine learning for the practicing anesthesiologist, describing how decision-making behaviors can emerge from simple equations. Relevant clinical questions are introduced to illustrate how machine learning might help solve them-perhaps bringing anesthesiology into an era of machine-assisted discovery.

Prediction of Bispectral Index during Target-controlled Infusion of Propofol and Remifentanil

Background

The discrepancy between predicted effect-site concentration and measured bispectral index is problematic during intravenous anesthesia with target-controlled infusion of propofol and remifentanil. We hypothesized that bispectral index during total intravenous anesthesia would be more accurately predicted by a deep learning approach.

Methods

Long short-term memory and the feed-forward neural network were sequenced to simulate the pharmacokinetic and pharmacodynamic parts of an empirical model, respectively, to predict intraoperative bispectral index during combined use of propofol and remifentanil. Inputs of long short-term memory were infusion histories of propofol and remifentanil, which were retrieved from target-controlled infusion pumps for 1,800 s at 10-s intervals. Inputs of the feed-forward network were the outputs of long short-term memory and demographic data such as age, sex, weight, and height. The final output of the feed-forward network was the bispectral index. The performance of bispectral index prediction was compared between the deep learning model and previously reported response surface model.

Results

The model hyperparameters comprised 8 memory cells in the long short-term memory layer and 16 nodes in the hidden layer of the feed-forward network. The model training and testing were performed with separate data sets of 131 and 100 cases. The concordance correlation coefficient (95% CI) were 0.561 (0.560 to 0.562) in the deep learning model, which was significantly larger than that in the response surface model (0.265 [0.263 to 0.266], P &lt; 0.001).

Conclusions

The deep learning model–predicted bispectral index during target-controlled infusion of propofol and remifentanil more accurately compared to the traditional model. The deep learning approach in anesthetic pharmacology seems promising because of its excellent performance and extensibility.

Intelligent framework for diagnosis of frozen shoulder using cross sectional survey and case studies

OBJECTIVES

Frozen shoulder is a disease in which shoulder becomes stiff. Accurate diagnosis of frozen shoulder is helpful in providing economical and effective treatment for patients. This research provides the classification of unstructured data using data mining techniques. Prediction results are validated by K-fold cross-validation method. It also provides accurate diagnosis of frozen shoulder using Naïve Bayesian and Random Forest models. At the end results are presented by performance measure techniques.

METHODS

In this research, 145 respondents (patients) with a severe finding of frozen shoulder are included. They are selected on premise of (clinical) assessment confirmed after by MRI. This data is taken from the department of Orthopedics (Pakistan Institute of Medical Sciences Islamabad and Railway Hospital Rawalpindi) between September 2014 to November 2015. Frozen shoulder is categorized on the basis of MRI result. The predictor variables are taken from patient survey and patient reports, which consisted of 35+ variables. The outcome variable is coded into numeric system of "intact" and "no-intact". The outcome variable is assigned into numeric code, 1 for "intact" and 0 for "no-intact". "Intact" group is used as an indication that tissue is damaged badly and "no-intact" is classified as normal. Distribution of result is 110 patients for "Intact" group and 35 patients for "No-Intact" group (false positive rate was 24 %). In this research we have utilized two methods i.e. Naive Bayes and Random Forest. A statistics regression model (Logistic regression) to categorize frozen shoulder finding into "intact" and "no-intact" classes. In the end, we validated our results by Bayesian theorem. This gives a rough estimate about the probability of frozen shoulder.

RESULTS

In this research, our anticipated and predictive procedures gave better outcome as compared to statistical techniques. The specificity and sensitivity ratio of predicting a frozen shoulder are better in the Naïve Bayes as compared to Random Forest. In end the likelihood ratio results are used with Bayesian theorem for final evaluation of the results, from this we conclude predictive model is valid model for classification of frozen shoulder.

CONCLUSIONS

We have used three predictive models in our study to classify frozen shoulder. Then we validated our predictive results by Bayesian theorem to give a rough estimate about the probability of occurrence of disease or not. This enhances the clinical decision making regarding frozen shoulder.

Predicting rotator cuff tears using data mining and Bayesian likelihood ratios

OBJECTIVES

Rotator cuff tear is a common cause of shoulder diseases. Correct diagnosis of rotator cuff tears can save patients from further invasive, costly and painful tests. This study used predictive data mining and Bayesian theory to improve the accuracy of diagnosing rotator cuff tears by clinical examination alone.

METHODS

In this retrospective study, 169 patients who had a preliminary diagnosis of rotator cuff tear on the basis of clinical evaluation followed by confirmatory MRI between 2007 and 2011 were identified. MRI was used as a reference standard to classify rotator cuff tears. The predictor variable was the clinical assessment results, which consisted of 16 attributes. This study employed 2 data mining methods (ANN and the decision tree) and a statistical method (logistic regression) to classify the rotator cuff diagnosis into "tear" and "no tear" groups. Likelihood ratio and Bayesian theory were applied to estimate the probability of rotator cuff tears based on the results of the prediction models.

RESULTS

Our proposed data mining procedures outperformed the classic statistical method. The correction rate, sensitivity, specificity and area under the ROC curve of predicting a rotator cuff tear were statistical better in the ANN and decision tree models compared to logistic regression. Based on likelihood ratios derived from our prediction models, Fagan's nomogram could be constructed to assess the probability of a patient who has a rotator cuff tear using a pretest probability and a prediction result (tear or no tear).

CONCLUSIONS

Our predictive data mining models, combined with likelihood ratios and Bayesian theory, appear to be good tools to classify rotator cuff tears as well as determine the probability of the presence of the disease to enhance diagnostic decision making for rotator cuff tears.

Application of gene expression programming and neural networks to predict adverse events of radical hysterectomy in cervical cancer patients

The aim of this article was to compare gene expression programming (GEP) method with three types of neural networks in the prediction of adverse events of radical hysterectomy in cervical cancer patients. One-hundred and seven patients treated by radical hysterectomy were analyzed. Each record representing a single patient consisted of 10 parameters. The occurrence and lack of perioperative complications imposed a two-class classification problem. In the simulations, GEP algorithm was compared to a multilayer perceptron (MLP), a radial basis function network neural, and a probabilistic neural network. The generalization ability of the models was assessed on the basis of their accuracy, the sensitivity, the specificity, and the area under the receiver operating characteristic curve (AUROC). The GEP classifier provided best results in the prediction of the adverse events with the accuracy of 71.96 %. Comparable but slightly worse outcomes were obtained using MLP, i.e., 71.87 %. For each of measured indices: accuracy, sensitivity, specificity, and the AUROC, the standard deviation was the smallest for the models generated by GEP classifier.

Development and evaluation of a simple and effective prediction approach for identifying those at high risk of dyslipidemia in rural adult residents

Background

Dyslipidemia is an extremely prevalent but preventable risk factor for cardiovascular disease. However, many dyslipidemia patients remain undetected in resource limited settings. The study was performed to develop and evaluate a simple and effective prediction approach without biochemical parameters to identify those at high risk of dyslipidemia in rural adult population.

Methods

Demographic, dietary and lifestyle, and anthropometric data were collected by a cross-sectional survey from 8,914 participants living in rural areas aged 35–78 years. There were 6,686 participants randomly selected into a training group for constructing the artificial neural network (ANN) and logistic regression (LR) prediction models. The remaining 2,228 participants were assigned to a validation group for performance comparisons of ANN and LR models. The predictors of dyslipidemia risk were identified from the training group using multivariate logistic regression analysis. Predictive performance was evaluated by receiver operating characteristic (ROC) curve.

Results

Some risk factors were significantly associated with dyslipidemia, including age, gender, educational level, smoking, high-fat diet, vegetable and fruit intake, family history, physical activity, and central obesity. For the ANN model, the sensitivity, specificity, positive and negative likelihood ratio, positive and negative predictive values were 90.41%, 76.66%, 3.87, 0.13, 76.33%, and 90.58%, respectively, while LR model were only 57.37%, 70.91%, 1.97, 0.60, 62.09%, and 66.73%, respectively. The area under the ROC cure (AUC) value of the ANN model was 0.86±0.01, showing more accurate overall performance than traditional LR model (AUC = 0.68±0.01, P<0.001).

Conclusion

The ANN model is a simple and effective prediction approach to identify those at high risk of dyslipidemia, and it can be used to screen undiagnosed dyslipidemia patients in rural adult population. Further work is planned to confirm these results by incorporating multi-center and longer follow-up data.

A hybrid decision support model to discover informative knowledge in diagnosing acute appendicitis

Background

The aim of this study is to develop a simple and reliable hybrid decision support model by combining statistical analysis and decision tree algorithms to ensure high accuracy of early diagnosis in patients with suspected acute appendicitis and to identify useful decision rules.

Methods

We enrolled 326 patients who attended an emergency medical center complaining mainly of acute abdominal pain. Statistical analysis approaches were used as a feature selection process in the design of decision support models, including the Chi-square test, Fisher's exact test, the Mann-Whitney U-test (p < 0.01), and Wald forward logistic regression (entry and removal criteria of 0.01 and 0.05, or 0.05 and 0.10, respectively). The final decision support models were constructed using the C5.0 decision tree algorithm of Clementine 12.0 after pre-processing.

Results

Of 55 variables, two subsets were found to be indispensable for early diagnostic knowledge discovery in acute appendicitis. The two subsets were as follows: (1) lymphocytes, urine glucose, total bilirubin, total amylase, chloride, red blood cell, neutrophils, eosinophils, white blood cell, complaints, basophils, glucose, monocytes, activated partial thromboplastin time, urine ketone, and direct bilirubin in the univariate analysis-based model; and (2) neutrophils, complaints, total bilirubin, urine glucose, and lipase in the multivariate analysis-based model. The experimental results showed that the model with univariate analysis (80.2%, 82.4%, 78.3%, 76.8%, 83.5%, and 80.3%) outperformed models using multivariate analysis (71.6%, 69.3%, 73.7%, 69.7%, 73.3%, and 71.5% with entry and removal criteria of 0.01 and 0.05; 73.5%, 66.0%, 80.0%, 74.3%, 72.9%, and 73.0% with entry and removal criteria of 0.05 and 0.10) in terms of accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and area under ROC curve, during a 10-fold cross validation. A statistically significant difference was detected in the pairwise comparison of ROC curves (p < 0.01, 95% CI, 3.13-14.5; p < 0.05, 95% CI, 1.54-13.1). The larger induced decision model was more effective for identifying acute appendicitis in patients with acute abdominal pain, whereas the smaller induced decision tree was less accurate with the test data.

Conclusions

The decision model developed in this study can be applied as an aid in the initial decision making of clinicians to increase vigilance in cases of suspected acute appendicitis.

Empirical antibiotic therapy (ABT) of lower respiratory tract infections (LRTI) in the elderly: application of artificial neural network (ANN). Preliminary results

LRTI are among the most common diseases in developed countries, including chronic obstructive pulmonary disease (COPD), one of the most frequent conditions. Their treatment in general practice is often unsuccessful and this increases hospital admissions. We know, bacterial infections in the elderly show a higher morbidity and mortality, either for more severe symptoms, than in younger adults, or because the causing agent often remains unknown. The need for a quick initiation of ABT often requires to chose on empirical grounds. To date there are no official guidelines for empirical ABT of COPD exacerbations, but only heterogeneous and often conflicting recommendations exist. The aim of our study was to identify a tool to guide the choice of the most effective empirical ABT when symptoms are acute and bacteriological tests cannot be performed. We used an ANN to study 117 patients aged between 55 and 97 years (mean 81.5 ± 8.7 years) (± S.D.), admitted with a diagnosis of pneumonia, COPD exacerbation or pneumonia with respiratory failure. We registered symptoms at onset and some individual variables such as age, sex, risk factors, comorbidity, current drug therapies. Then the ANN was applied to choose ABT in 20 patients versus 20 subjects whose therapy was chosen by the physicians, comparing these groups for therapy's efficacy, mean durations of therapy and hospitalization (H). In the learning phase, the ANN could predict the resolution index 99.05% of the time (i.e., 104 times) with a ± S.D. = 0.23. After the training, during the test phase, the network predicted the resolution index 91.67% of the time (i.e., 11 times) with a ± S.D. = 0.54, thus proving the validity of the relations identified during the learning phase. Preliminary results of the application of our tool, show the ANN allowed us to greatly reduce the duration of the ABT and subsequently of the H. Based on preliminary results, we assume that the use of ANN can make a valuable contribution in the choice of empirical ABT in the course of acute lung diseases in elderly.