RESUMEN
Among patients with epilepsy, 30-40% experience recurrent seizures even after adequate antiseizure medications therapies, making them refractory. The early identification of refractory epilepsy is important to provide timely surgical treatment for these patients. In this study, we analyze interictal electroencephalography (EEG) data to predict drug refractoriness in patients with temporal lobe epilepsy (TLE) who were treated with monotherapy at the time of the first EEG acquisition. Various EEG features were extracted, including statistical measurements and interchannel coherence. Feature selection was performed to identify the optimal features, and classification was conducted using different classifiers. Functional connectivity and graph theory measurements were calculated to identify characteristics of refractory TLE. Among the 48 participants, 34 (70.8%) were responsive, while 14 (29.2%) were refractory over a mean follow-up duration of 38.5 months. Coherence feature within the gamma frequency band exhibited the most favorable performance. The light gradient boosting model, employing the mutual information filter-based feature selection method, demonstrated the highest performance (AUROC = 0.821). Compared to the responsive group, interchannel coherence displayed higher values in the refractory group. Interestingly, graph theory measurements using EEG coherence exhibited higher values in the refractory group than in the responsive group. Our study has demonstrated a promising method for the early identification of refractory TLE utilizing machine learning algorithms.
Asunto(s)
Anticonvulsivantes , Electroencefalografía , Epilepsia del Lóbulo Temporal , Humanos , Epilepsia del Lóbulo Temporal/tratamiento farmacológico , Epilepsia del Lóbulo Temporal/fisiopatología , Epilepsia del Lóbulo Temporal/cirugía , Femenino , Masculino , Adulto , Anticonvulsivantes/uso terapéutico , Persona de Mediana Edad , Epilepsia Refractaria/tratamiento farmacológico , Epilepsia Refractaria/fisiopatología , Adulto JovenRESUMEN
The ongoing coronavirus disease 2019 (COVID-19) pandemic presents serious public health threats. Omicron, the current most prevalent strain of COVID-19, has a low fatality rate and very high transmissibility, so the number of patients with mild symptoms of COVID-19 is rapidly increasing. This change of pandemic challenges medical systems worldwide in many aspects, including sharp increases in demands for hospital infrastructure, critical shortages in medical equipment, and medical staff. Predicting deterioration in mild patients could alleviate these problems. A novel scoring system was proposed for predicting the deterioration of patients whose condition may worsen rapidly and those who all still mild or asymptomatic. Retrospective cohorts of 954 and 2,035 patients that quarantined in the Residential Treatment Center were assembled for derivation and external validation of mild COVID-19, respectively. Deterioration was defined as transfer to a local hospital due to worsening condition of the patients during the 2-week isolation period. A total of 15 variables: sex, age, seven pre-existing conditions (diabetes, hypertension, cardiovascular disease, respiratory disease, liver disease, kidney disease, and organ transplant), and five vital signs (systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR), body temperature, and oxygen saturation (SpO2)) were collected. A scoring system was developed using seven variables (age, pulse rate, SpO2, SBP, DBP, temperature, and hypertension) with significant differences between the transfer and not transfer groups in logistic regression. The proposed system was compared with existing scoring systems that assess the severity of patient conditions. The performance of the proposed scoring system to predict deterioration in patients with mild COVID-19 showed an area under the receiver operating characteristic (AUC) of 0.868. This is a statistically significant improvement compared to the performance of the previous patient condition assessment scoring systems. During external validation, the proposed system showed the best and most robust predictive performance (AUC = 0.768; accuracy = 0.899). In conclusion, we proposed a novel scoring system for predicting patients with mild COVID-19 who will experience deterioration which could predict the deterioration of the patient's condition early with high predictive performance. Furthermore, because the scoring system does not require special calculations, it can be easily measured to predict the deterioration of a patients' condition. This system can be used as effective tool for early detection of deterioration in mild COVID-19 patients.
Asunto(s)
COVID-19 , SARS-CoV-2 , Humanos , COVID-19/epidemiología , COVID-19/diagnóstico , Masculino , Femenino , Estudios Retrospectivos , Persona de Mediana Edad , República de Corea/epidemiología , Anciano , SARS-CoV-2/aislamiento & purificación , Adulto , Índice de Severidad de la Enfermedad , PandemiasRESUMEN
OBJECTIVE: Failure to receive prompt blood transfusion leads to severe complications if massive bleeding occurs during surgery. For the timely preparation of blood products, predicting the possibility of massive transfusion (MT) is essential to decrease morbidity and mortality. This study aimed to develop a model for predicting MT 10 min in advance using non-invasive bio-signal waveforms that change in real-time. METHODS: In this retrospective study, we developed a deep learning-based algorithm (DLA) to predict intraoperative MT within 10 min. MT was defined as the transfusion of 3 or more units of red blood cells within an hour. The datasets consisted of 18,135 patients who underwent surgery at Seoul National University Hospital (SNUH) for model development and internal validation and 621 patients who underwent surgery at the Boramae Medical Center (BMC) for external validation. We constructed the DLA by using features extracted from plethysmography (collected at 500 Hz) and hematocrit measured during surgery. RESULTS: Among 18,135 patients in SNUH and 621 patients in BMC, 265 patients (1.46%) and 14 patients (2.25%) received MT during surgery, respectively. The area under the receiver operating characteristic curve (AUROC) of DLA predicting intraoperative MT before 10 min was 0.962 (95% confidence interval [CI], 0.948-0.974) in internal validation and 0.922 (95% CI, 0.882-0.959) in external validation, respectively. CONCLUSION: The DLA can successfully predict intraoperative MT using non-invasive bio-signal waveforms.
Asunto(s)
Transfusión Sanguínea , Humanos , Femenino , Masculino , Estudios Retrospectivos , Persona de Mediana Edad , Algoritmos , Anciano , Monitoreo Intraoperatorio/métodos , Monitorización Hemodinámica/métodos , Adulto , Aprendizaje Profundo , Curva ROC , Hemodinámica , Hematócrito , Pérdida de Sangre QuirúrgicaRESUMEN
Background: Current risk stratification strategies for patients with hypertrophic cardiomyopathy (HCM) are limited to traditional methodologies. Objectives: The authors aimed to establish machine learning (ML)-based models to discriminate major cardiovascular events in patients with HCM. Methods: We enrolled consecutive HCM patients from 2 tertiary referral centers and used 25 clinical and echocardiographic features to discriminate major adverse cardiovascular events (MACE), including all-cause death, admission for heart failure (HF-adm), and stroke. The best model was selected for each outcome using the area under the receiver operating characteristic curve (AUROC) with 20-fold cross-validation. After testing in the external validation cohort, the relative importance of features in discriminating each outcome was determined using the SHapley Additive exPlanations (SHAP) method. Results: In total, 2,111 patients with HCM (age 61.4 ± 13.6 years; 67.6% men) were analyzed. During the median 4.0 years of follow-up, MACE occurred in 341 patients (16.2%). Among the 4 ML models, the logistic regression model achieved the best AUROC of 0.800 (95% CI: 0.760-0.841) for MACE, 0.789 (95% CI: 0.736-0.841) for all-cause death, 0.798 (95% CI: 0.736-0.860) for HF-adm, and 0.807 (95% CI: 0.754-0.859) for stroke. The discriminant ability of the logistic regression model remained excellent when applied to the external validation cohort for MACE (AUROC = 0.768), all-cause death (AUROC = 0.750), and HF-adm (AUROC = 0.806). The SHAP analysis identified left atrial diameter and hypertension as important variables for all outcomes of interest. Conclusions: The proposed ML models incorporating various phenotypes from patients with HCM accurately discriminated adverse cardiovascular events and provided variables with high importance for each outcome.
RESUMEN
BACKGROUND: Falls impact over 25% of older adults annually, making fall prevention a critical public health focus. We aimed to develop and validate a machine learning-based prediction model for serious fall-related injuries (FRIs) among community-dwelling older adults, incorporating various medication factors. METHODS: Utilizing annual national patient sample data, we segmented outpatient older adults without FRIs in the preceding three months into development and validation cohorts based on data from 2018 and 2019, respectively. The outcome of interest was serious FRIs, which we defined operationally as incidents necessitating an emergency department visit or hospital admission, identified by the diagnostic codes of injuries that are likely associated with falls. We developed four machine-learning models (light gradient boosting machine, Catboost, eXtreme Gradient Boosting, and Random forest), along with a logistic regression model as a reference. RESULTS: In both cohorts, FRIs leading to hospitalization/emergency department visits occurred in approximately 2% of patients. After selecting features from initial set of 187, we retained 26, with 15 of them being medication-related. Catboost emerged as the top model, with area under the receiver operating characteristic of 0.700, along with sensitivity and specificity rates around 65%. The high-risk group showed more than threefold greater risk of FRIs than the low-risk group, and model interpretations aligned with clinical intuition. CONCLUSION: We developed and validated an explainable machine-learning model for predicting serious FRIs in community-dwelling older adults. With prospective validation, this model could facilitate targeted fall prevention strategies in primary care or community-pharmacy settings.
Asunto(s)
Vida Independiente , Aprendizaje Automático , Humanos , Anciano , Factores de Riesgo , República de Corea/epidemiologíaRESUMEN
Postoperative desaturation is a common post-surgery pulmonary complication. The real-time prediction of postoperative desaturation can become a preventive measure, and real-time changes in spirometry data can provide valuable information on respiratory mechanics. However, there is a lack of related research, specifically on using spirometry signals as inputs to machine learning (ML) models. We developed an ML model and postoperative desaturation prediction index (DPI) by analyzing intraoperative spirometry signals in patients undergoing laparoscopic surgery. We analyzed spirometry data from patients who underwent laparoscopic, robot-assisted gynecologic, or urologic surgery, identifying postoperative desaturation as a peripheral arterial oxygen saturation level below 95%, despite facial oxygen mask usage. We fitted the ML model on two separate datasets collected during different periods. (Datasets A and B). Dataset A (Normal 133, Desaturation 74) was used for the entire experimental process, including ML model fitting, statistical analysis, and DPI determination. Dataset B (Normal 20, Desaturation 4) was only used for verify the ML model and DPI. Four feature categories-signal property, inter-/intra-position correlation, peak value/interval variability, and demographics-were incorporated into the ML models via filter and wrapper feature selection methods. In experiments, the ML model achieved an adequate predictive capacity for postoperative desaturation, and the performance of the DPI was unbiased.
Asunto(s)
Oximetría , Oxígeno , Humanos , Femenino , Oximetría/métodos , Complicaciones Posoperatorias , Mecánica Respiratoria , EspirometríaRESUMEN
Coronavirus has caused many casualties and is still spreading. Some people experience rapid deterioration that is mild at first. The aim of this study is to develop a deterioration prediction model for mild COVID-19 patients during the isolation period. We collected vital signs from wearable devices and clinical questionnaires. The derivation cohort consisted of people diagnosed with COVID-19 between September and December 2021, and the external validation cohort collected between March and June 2022. To develop the model, a total of 50 participants wore the device for an average of 77 h. To evaluate the model, a total of 181 infected participants wore the device for an average of 65 h. We designed machine learning-based models that predict deterioration in patients with mild COVID-19. The prediction model, 10 min in advance, showed an area under the receiver characteristic curve (AUC) of 0.99, and the prediction model, 8 h in advance, showed an AUC of 0.84. We found that certain variables that are important to model vary depending on the point in time to predict. Efficient deterioration monitoring in many patients is possible by utilizing data collected from wearable sensors and symptom self-reports.
Asunto(s)
COVID-19 , Dispositivos Electrónicos Vestibles , Humanos , Autoinforme , Encuestas y Cuestionarios , Aprendizaje AutomáticoRESUMEN
Epilepsy is a neurological disorder in which the brain is transiently altered. Predicting outcomes in epilepsy is essential for providing feedback that can foster improved outcomes in the future. This study aimed to investigate whether applying spectral and temporal filters to resting-state electroencephalography (EEG) signals could improve the prediction of outcomes for patients taking antiseizure medication to treat temporal lobe epilepsy (TLE). We collected EEG data from a total of 46 patients (divided into a seizure-free group (SF, n = 22) and a non-seizure-free group (NSF, n = 24)) with TLE and retrospectively reviewed their clinical data. We segmented spectral and temporal ranges with various time-domain features (Hjorth parameters, statistical parameters, energy, zero-crossing rate, inter-channel correlation, inter-channel phase locking value and spectral information derived from Fourier transform, Stockwell transform, and wavelet transform) and compared their performance by applying an optimal frequency strategy, an optimal duration strategy, and a combination strategy. For all time-domain features, the optimal frequency and time combination strategy showed the highest performance in distinguishing SF patients from NSF patients (area under the curve (AUC) = 0.790 ± 0.159). Furthermore, optimal performance was achieved by utilizing a feature vector derived from statistical parameters within the 39- to 41-Hz frequency band with a window length of 210 s, as evidenced by an AUC of 0.748. By identifying the optimal parameters, we improved the performance of the prediction model. These parameters can serve as standard parameters for predicting outcomes based on resting-state EEG signals.
Asunto(s)
Epilepsia del Lóbulo Temporal , Epilepsia , Humanos , Epilepsia del Lóbulo Temporal/tratamiento farmacológico , Estudios Retrospectivos , Electroencefalografía , Aprendizaje AutomáticoRESUMEN
Older adults are more likely to require emergency department (ED) visits than others, which might be attributed to their medication use. Being able to predict the likelihood of an ED visit using prescription information and readily available data would be useful for primary care. This study aimed to predict the likelihood of ED visits using extensive medication variables generated according to explicit clinical criteria for elderly people and high-risk medication categories by applying machine learning (ML) methods. Patients aged ≥ 65 years were included, and ED visits were predicted with 146 variables, including demographic and comprehensive medication-related factors, using nationwide claims data. Among the eight ML models, the final model was developed using LightGBM, which showed the best performance. The final model incorporated 93 predictors, including six sociodemographic, 28 comorbidity, and 59 medication-related variables. The final model had an area under the receiver operating characteristic curve of 0.689 in the validation cohort. Approximately half of the top 20 strong predictors were medication-related variables. Here, an ED visit risk prediction model for older people was developed and validated using administrative data that can be easily applied in clinical settings to screen patients who are likely to visit an ED.
Asunto(s)
Servicio de Urgencia en Hospital , Vida Independiente , Anciano , Humanos , Comorbilidad , Aprendizaje Automático , Curva ROC , Estudios RetrospectivosRESUMEN
BACKGROUND AND OBJECTIVE: Neuromuscular disorders are diseases that damage our ability to control body movements. Needle electromyography (nEMG) is often used to diagnose neuromuscular disorders, which is an electrophysiological test measuring electric signals generated from a muscle using an invasive needle. Characteristics of nEMG signals are manually analyzed by an electromyographer to diagnose the types of neuromuscular disorders, and this process is highly dependent on the subjective experience of the electromyographer. Contemporary computer-aided methods utilized deep learning image classification models to classify nEMG signals which are not optimized for classifying signals. Additionally, model explainability was not addressed which is crucial in medical applications. This study aims to improve prediction accuracy, inference time, and explain model predictions in nEMG neuromuscular disorder classification. METHODS: This study introduces the nEMGNet, a one-dimensional convolutional neural network with residual connections designed to extract features from raw signals with higher accuracy and faster speed compared to image classification models from previous works. Next, the divide-and-vote (DiVote) algorithm was designed to integrate each subject's heterogeneous nEMG signal data structures and to utilize muscle subtype information for higher accuracy. Finally, feature visualization was used to identify the causality of nEMGNet diagnosis predictions, to ensure that nEMGNet made predictions on valid features, not artifacts. RESULTS: The proposed method was tested using 376 nEMG signals measured from 57 subjects between June 2015 to July 2020 in Seoul National University Hospital. The results from the three-class classification task demonstrated that nEMGNet's prediction accuracy of nEMG signal segments was 62.35%, and the subject diagnosis prediction accuracy of nEMGNet and the DiVote algorithm was 83.69 %, over 5-fold cross-validation. nEMGNet outperformed all models from previous works on nEMG diagnosis classification, and heuristic analysis of feature visualization results indicate that nEMGNet learned relevant nEMG signal characteristics. CONCLUSIONS: This study introduced nEMGNet and DiVote algorithm which demonstrated fast and accurate performance in predicting neuromuscular disorders based on nEMG signals. The proposed method may be applied in medicine to support real-time electrophysiologic diagnosis.
Asunto(s)
Algoritmos , Redes Neurales de la Computación , Humanos , Electromiografía/métodos , MovimientoRESUMEN
Currently, the identification of stroke patients with an increased suicide risk is mainly based on self-report questionnaires, and this method suffers from a lack of objectivity. This study developed and validated a suicide ideation (SI) prediction model using clinical data and identified SI predictors. Significant variables were selected through traditional statistical analysis based on retrospective data of 385 stroke patients; the data were collected from October 2012 to March 2014. The data were then applied to three boosting models (Xgboost, CatBoost, and LGBM) to identify the comparative and best performing models. Demographic variables that showed significant differences between the two groups were age, onset, type, socioeconomic, and education level. Additionally, functional variables also showed a significant difference with regard to ADL and emotion (p < 0.05). The CatBoost model (0.900) showed higher performance than the other two models; and depression, anxiety, self-efficacy, and rehabilitation motivation were found to have high importance. Negative emotions such as depression and anxiety showed a positive relationship with SI and rehabilitation motivation and self-efficacy displayed an inverse relationship with SI. Machine learning-based SI models could augment SI prevention by helping rehabilitation and medical professionals identify high-risk stroke patients in need of SI prevention intervention.
Asunto(s)
Ansiedad , Ideación Suicida , Trastornos de Ansiedad , Humanos , Aprendizaje Automático , Estudios Retrospectivos , Factores de RiesgoRESUMEN
Twenty-five cadaveric adult femora's anteversion angles were measured to develop a highly efficient and reproducible femoral anteversion measurement method using computed tomography (CT). Digital photography captured the proximal femur's two reference lines, head-to-neck (H-N) and head-to-greater trochanter (H-G). Six reference lines (A/B in transverse section; C, axial oblique section; D/E, conventional 3D reconstruction; and M, volumetric 3D reconstruction) from CT scans were used. The posterior condylar line was used as a distal femoral reference. As measured with the H-N and H-G lines, the anteversion means were 10.43° and 19.50°, respectively. Gross anteversion measured with the H-G line had less interobserver bias (ICC; H-N = 0.956, H-G = 0.982). The 2D transverse and volumetric 3D CT sections' B/M lines were consistent with the H-N line (p: B = 0.925, M = 0.122) and the 2D axial oblique section's C line was consistent with the H-G line (p < 0.1). The D/E lines differed significantly from the actual gross images (p < 0.05). Among several CT scan femoral anteversion measurement methods, the novel anteversion angle measurement method using CT scans' axial oblique section was approximated with actual gross femoral anteversion angle from the femoral head to the greater trochanter.
RESUMEN
Background: Depression and suicide are critical social problems worldwide, but tools to objectively diagnose them are lacking. Therefore, this study aimed to diagnose depression through machine learning and determine whether it is possible to identify groups at high risk of suicide through words spoken by the participants in a semi-structured interview. Methods: A total of 83 healthy and 83 depressed patients were recruited. All participants were recorded during the Mini-International Neuropsychiatric Interview. Through the suicide risk assessment from the interview items, participants with depression were classified into high-suicide-risk (31 participants) and low-suicide-risk (52 participants) groups. The recording was transcribed into text after only the words uttered by the participant were extracted. In addition, all participants were evaluated for depression, anxiety, suicidal ideation, and impulsivity. The chi-square test and student's T-test were used to compare clinical variables, and the Naive Bayes classifier was used for the machine learning text model. Results: A total of 21,376 words were extracted from all participants and the model for diagnosing patients with depression based on this text confirmed an area under the curve (AUC) of 0.905, a sensitivity of 0.699, and a specificity of 0.964. In the model that distinguished the two groups using statistically significant demographic variables, the AUC was only 0.761. The DeLong test result (p-value 0.001) confirmed that the text-based classification was superior to the demographic model. When predicting the high-suicide-risk group, the demographics-based AUC was 0.499, while the text-based one was 0.632. However, the AUC of the ensemble model incorporating demographic variables was 0.800. Conclusion: The possibility of diagnosing depression using interview text was confirmed; regarding suicide risk, the diagnosis accuracy increased when demographic variables were incorporated. Therefore, participants' words during an interview show significant potential as an objective and diagnostic marker through machine learning.
RESUMEN
INTRODUCTION: Parkinson's disease (PD) is a neurodegenerative disorder with only symptomatic treatments currently available. Although correct, early diagnoses of PD are important, the existing diagnostic method based on pathologic examinations only has an accuracy of approximately 80.6%. Although electroencephalography (EEG)-based assistive technology has been introduced, it has been difficult to implement in practice due to the high computational complexity and low accuracy of the analysis methods. This study proposed a fast, accurate PD prediction method using the Hjorth parameter and the gradient boosting decision tree (GBDT) algorithm. METHOD: We used an open EEG dataset with 41 PD patients and 41 healthy controls (HCs); EEG signals were recorded from participants at the University of New Mexico (PD: 27 vs. HC: 27) and University of Iowa (PD: 14 vs. HC: 14). We explored the analytic time segment and frequency range in which the Hjorth parameter best represents the EEG characteristics of PD patients. RESULTS: Our best model (CatBoost-based) distinguished PD patients from controls with an accuracy of 89.3%, an area under the receiver operating characteristics curve (AUC) of 0.912, an F-score of 0.903, and an odds ratio of 115.5. These results showed that our models outperformed those of all other previous works and were even superior to previously known pathologic examination-based diagnoses with long-term follow-up (accuracy = 83.9%). CONCLUSION: The proposed methods are expected to be utilized as an effective method for improving the diagnosis of PD.
Asunto(s)
Enfermedad de Parkinson , Algoritmos , Árboles de Decisión , Electroencefalografía/métodos , Humanos , Enfermedad de Parkinson/diagnóstico , Enfermedad de Parkinson/patologíaRESUMEN
Impaired cerebral autoregulation (CA) can cause negative outcomes in neurological conditions. Real-time CA monitoring can predict and thereby help prevent postoperative complications for neurosurgery patients, especially those suffering from moyamoya disease (MMD). We applied the concept of moving average to the correlation between mean arterial blood pressure (MBP) and cerebral oxygen saturation (SCO2) to monitor CA in real time, revealing optimal window size for the moving average. The experiment was conducted with 68 surgical vital-sign records containing MBP and SCO2. To evaluate CA, the cerebral oximetry index (COx) and coherence obtained from transfer function analysis (TFA) were calculated and compared between patients with postoperative infarction and those who without. For real-time monitoring, the moving average was applied to COx and coherence to determine the differences between groups, and the optimal moving-average window size was identified. The average COx and coherence within the very-low-frequency (VLF) range (0.02-0.07 Hz) during the entire surgery were significantly different between the groups (COx: AUROC = 0.78, p = 0.003; coherence: AUROC = 0.69, p = 0.029). For the case of real-time monitoring, COx showed a reasonable performance (AUROC > 0.74) with moving-average window sizes larger than 30 minutes. Coherence showed an AUROC > 0.7 for time windows of up to 60 minutes; however, for windows larger than this threshold, the performance became unstable. With an appropriate window size, COx showed stable performance as a predictor of postoperative infarction in MMD patients.
Asunto(s)
Enfermedad de Moyamoya , Humanos , Enfermedad de Moyamoya/cirugía , Oximetría/métodos , Monitoreo Intraoperatorio/métodos , Espectroscopía Infrarroja Corta/métodos , Circulación Cerebrovascular/fisiología , Puente Cardiopulmonar , Homeostasis/fisiologíaRESUMEN
OBJECTIVE: Monitoring intracranial and arterial blood pressure (ICP and ABP, respectively) provides crucial information regarding the neurological status of patients with traumatic brain injury (TBI). However, these signals are often heavily affected by artifacts, which may significantly reduce the reliability of the clinical determinations derived from the signals. The goal of this work was to eliminate signal artifacts from continuous ICP and ABP monitoring via deep learning techniques and to assess the changes in the prognostic capacities of clinical parameters after artifact elimination. METHODS: The first 24 hours of monitoring ICP and ABP in a total of 309 patients with TBI was retrospectively analyzed. An artifact elimination model for ICP and ABP was constructed via a stacked convolutional autoencoder (SCAE) and convolutional neural network (CNN) with 10-fold cross-validation tests. The prevalence and prognostic capacity of ICP- and ABP-related clinical events were compared before and after artifact elimination. RESULTS: The proposed SCAE-CNN model exhibited reliable accuracy in eliminating ABP and ICP artifacts (net prediction rates of 97% and 94%, respectively). The prevalence of ICP- and ABP-related clinical events (i.e., systemic hypotension, intracranial hypertension, cerebral hypoperfusion, and poor cerebrovascular reactivity) all decreased significantly after artifact removal. CONCLUSIONS: The SCAE-CNN model can be reliably used to eliminate artifacts, which significantly improves the reliability and efficacy of ICP- and ABP-derived clinical parameters for prognostic determinations after TBI.
RESUMEN
BACKGROUND: Hemodynamic instability and cardiovascular events heavily affect the prognosis of traumatic brain injury. Physiological signals are monitored to detect these events. However, the signals are often riddled with faulty readings, which jeopardize the reliability of the clinical parameters obtained from the signals. A machine-learning model for the elimination of artifactual events shows promising results for improving signal quality. However, the actual impact of the improvements on the performance of the clinical parameters after the elimination of the artifacts is not well studied. MATERIALS AND METHODS: The arterial blood pressure of 99 subjects with traumatic brain injury was continuously measured for 5 consecutive days, beginning on the day of admission. The machine-learning deep belief network was constructed to automatically identify and remove false incidences of hypotension, hypertension, bradycardia, tachycardia, and alterations in cerebral perfusion pressure (CPP). RESULTS: The prevalences of hypotension and tachycardia were significantly reduced by 47.5% and 13.1%, respectively, after suppressing false incidents (P=0.01). Hypotension was particularly effective at predicting outcome favorability and mortality after artifact elimination (P=0.015 and 0.027, respectively). In addition, increased CPP was also statistically significant in predicting outcomes (P=0.02). CONCLUSIONS: The prevalence of false incidents due to signal artifacts can be significantly reduced using machine-learning. Some clinical events, such as hypotension and alterations in CPP, gain particularly high predictive capacity for patient outcomes after artifacts are eliminated from physiological signals.