Precision medicine for traumatic cervical spinal cord injuries: accessible and interpretable machine learning models to predict individualized in-hospital outcomes

Early Prognostication of Critical Patients With Spinal Cord Injury: A Machine Learning Study With 1485 Cases

STUDY DESIGN

A retrospective case-series.

OBJECTIVE

The study aims to use machine learning to predict the discharge destination of spinal cord injury (SCI) patients in the intensive care unit.

SUMMARY OF BACKGROUND DATA

Prognostication following SCI is vital, especially for critical patients who need intensive care.

PATIENTS AND METHODS

Clinical data of patients diagnosed with SCI were extracted from a publicly available intensive care unit database. The first recorded data of the included patients were used to develop a total of 98 machine learning classifiers, seeking to predict discharge destination (eg, death, further medical care, home, etc.). The microaverage area under the curve (AUC) was the main indicator to assess discrimination. The best average-AUC classifier and the best death-sensitivity classifier were integrated into an ensemble classifier. The discrimination of the ensemble classifier was compared with top death-sensitivity classifiers and top average-AUC classifiers. In addition, prediction consistency and clinical utility were also assessed.

RESULTS

A total of 1485 SCI patients were included. The ensemble classifier had a microaverage AUC of 0.851, which was only slightly inferior to the best average-AUC classifier ( P =0.10). The best average-AUC classifier death sensitivity was much lower than that of the ensemble classifier. The ensemble classifier had a death sensitivity of 0.452, which was inferior to the top 8 death-sensitivity classifiers, whose microaverage AUC were inferior to the ensemble classifier ( P <0.05). In addition, the ensemble classifier demonstrated a comparable Brier score and superior net benefit in the DCA when compared with the performance of the origin classifiers.

CONCLUSIONS

The ensemble classifier shows an overall superior performance in predicting discharge destination, considering discrimination ability, prediction consistency, and clinical utility. This classifier system may aid in the clinical management of critical SCI patients in the early phase following injury.

LEVEL OF EVIDENCE

Level 3.

Sensitivity and specificity of machine learning and deep learning algorithms in the diagnosis of thoracolumbar injuries resulting in vertebral fractures: A systematic review and meta-analysis

Introduction

Clinicians encounter challenges in promptly diagnosing thoracolumbar injuries (TLIs) and fractures (VFs), motivating the exploration of Artificial Intelligence (AI) and Machine Learning (ML) and Deep Learning (DL) technologies to enhance diagnostic capabilities. Despite varying evidence, the noteworthy transformative potential of AI in healthcare, leveraging insights from daily healthcare data, persists.

Research question

This review investigates the utilization of ML and DL in TLIs causing VFs.

Materials and methods

Employing Preferred Reporting Items for Systematic Reviews and Meta-Analyzes (PRISMA) methodology, a systematic review was conducted in PubMed and Scopus databases, identifying 793 studies. Seventeen were included in the systematic review, and 11 in the meta-analysis. Variables considered encompassed publication years, geographical location, study design, total participants (14,524), gender distribution, ML or DL methods, specific pathology, diagnostic modality, test analysis variables, validation details, and key study conclusions. Meta-analysis assessed specificity, sensitivity, and conducted hierarchical summary receiver operating characteristic curve (HSROC) analysis.

Results

Predominantly conducted in China (29.41%), the studies involved 14,524 participants. In the analysis, 11.76% (N = 2) focused on ML, while 88.24% (N = 15) were dedicated to deep DL. Meta-analysis revealed a sensitivity of 0.91 (95% CI = 0.86-0.95), consistent specificity of 0.90 (95% CI = 0.86-0.93), with a false positive rate of 0.097 (95% CI = 0.068-0.137).

Conclusion

The study underscores consistent specificity and sensitivity estimates, affirming the diagnostic test's robustness. However, the broader context of ML applications in TLIs emphasizes the critical need for standardization in methodologies to enhance clinical utility.

Machine learning predictive models for grading bronchopulmonary dysplasia: umbilical cord blood IL-6 as a biomarker

Objectives

This study aimed to analyze the predictive value of umbilical cord blood Interleukin-6 (UCB IL-6) for the severity-graded BPD and to establish machine learning (ML) predictive models in a Chinese population based on the 2019 NRN evidence-based guidelines.

Methods

In this retrospective analysis, we included infants born with gestational age <32 weeks, who underwent UCB IL-6 testing within 24 h of admission to our NICU between 2020 and 2022. We collected their medical information encompassing the maternal, perinatal, and early neonatal phases. Furthermore, we classified the grade of BPD according to the 2019 NRN evidence-based guidelines. The correlation between UCB IL-6 and the grades of BPD was analyzed. Univariate analysis and ordinal logistic regression were employed to identify risk factors, followed by the development of ML predictive models based on XGBoost, CatBoost, LightGBM, and Random Forest. The AUROC was used to evaluate the diagnostic value of each model. Besides, we generated feature importance distribution plots based on SHAP values to emphasize the significance of UCB IL-6 in the models.

Results

The study ultimately enrolled 414 preterm infants, with No BPD group (n = 309), Grade 1 BPD group (n = 73), and Grade 2-3 BPD group (n = 32). The levels of UCB IL-6 increased with the grades of BPD. UCB IL-6 demonstrated clinical significance in predicting various grades of BPD, particularly in distinguishing Grade 2-3 BPD patients, with an AUROC of 0.815 (95% CI: 0.753-0.877). All four ML models, XGBoost, CatBoost, LightGBM, and Random Forest, exhibited Micro-average AUROC values of 0.841, 0.870, 0.851, and 0.878, respectively. Notably, UCB IL-6 consistently appeared as the most prominent feature across the feature importance distribution plots in all four models.

Conclusion

UCB IL-6 significantly contributes to predicting severity-graded BPD, especially in grade 2-3 BPD. Through the development of four ML predictive models, we highlighted UCB IL-6's importance.