Machine Learning Introduces Electrophysiology Assessment as the Best Predictor for the Recovery Prognosis of Spinal Cord Injury Patients for Personalized Rehabilitation Approaches

Abstract

The strong correlation between evoked potentials (EPs) and American Spinal Injury Association (ASIA) scores in individuals with spinal cord injury (SCI) suggests that EPs may serve as reliable predictive markers for rehabilitation progress. Numerous studies have confirmed a relationship between variations in somatosensory evoked potentials (SSEPs) and ASIA scores, especially in the early stages of SCI. Machine learning’s (ML’s) increasing importance in medicine is driven by the growing availability of health data and improved algorithms. It enables the creation of predictive models for disease diagnosis, progression prediction, personalized treatment, and improved healthcare efficiency. Data-driven approaches can significantly improve patient care, reduce costs, and facilitate personalized medicine. The meticulous analysis of medical data is crucial for timely disease identification, leading to effective symptom management and appropriate treatment. This study applies artificial intelligence to identify predictors of SCI progression, as measured by the disability index, ASIA impairment scale (AIS), and final motor recovery. We aim to clarify the prognostic role of electrophysiological testing (SSEPs, MEPs, and nerve conduction studies (NCSs)) in SCI. We analyzed data from a medical database of 123 records. We developed an ML-based intelligent system, utilizing ensemble algorithms combining decision trees and neural network approaches, to predict SCI recovery. Our evaluation showed SEP accuracies of 90% for motor recovery prediction and 80% for AIS scale determination, comparable to full electrophysiology evaluation accuracies of 93% and 89%, respectively, and generally superior results compared to MEP and NCS results. EPs emerged as the best predictors, comparable to a comprehensive electrophysiology assessment, significantly improving accuracy compared to clinical findings alone. An electrophysiological assessment, when available, increased overall accuracy for final motor recovery prediction to 93% (from a maximum of 75%) and, for ASIA score determination, to 89% (from a maximum of 66%). Further validation is needed with a larger dataset. Future research should validate that sensory electrophysiology assessment is a less expensive, portable, and simpler alternative to other prognostic tests and more effective than clinical assessments, like the AIS, biomarker for SCI, and personalized rehabilitation planning.

1. Introduction

A spinal cord injury (SCI) is damage to the nerve bundle transmitting signals between the brain and the body. The spinal cord, an extension of the central nervous system (CNS), begins at the medulla oblongata and ends at the conus medullaris, followed by the cauda equina. The brain controls bodily functions, while the spinal cord facilitates communication. These signals control muscle movement, sensation, and autonomic functions, like breathing and heart rate [1]. SCI can result from direct trauma or damage to surrounding tissues and vertebrae, causing temporary or permanent impairment below the injury site [2]. Symptoms depend on injury location and severity; higher injuries may cause tetraplegia, while lower injuries cause paraplegia. Paralysis can be immediate (primary damage) or progressive (secondary damage) due to bleeding, swelling, or cell death. The extent of nerve fiber damage affects recovery potential. Symptoms include numbness, pain, weakness, walking difficulties, and loss of bladder or bowel control. Breathing difficulties and altered sexual function may also occur [3].

1.1. Pathophysiology and Diagnosis of SCI

After an SCI, pathological changes affect recovery. Initial vascular disruption causes hypoperfusion and ischemia, leading to neuronal cell death. Inflammatory responses, oxidative stress, and apoptotic pathways contribute to tissue damage. Neurogenic shock (hypotension and bradycardia) may occur due to autonomic dysfunction. Tissue loss leads to cavity formation, complicating recovery [4]. Diagnosing SCI involves clinical assessments and imaging. Emergency evaluation includes testing movement, sensation, breathing, responsiveness, and muscle strength. Magnetic resonance imaging (MRI) detects spinal trauma, herniated discs, vascular abnormalities, and ligament damage [5]. Computerized tomography (CT) scans identify fractures, bleeding, and spinal stenosis [6]. X-rays quickly identify vertebral misalignment or fractures, aiding immediate decision-making [7].

1.2. Classification and Assessment of SCI

SCI classification uses anatomical and functional criteria. The spine is divided into five sections: cervical (C1–C7), thoracic (T1–T12), lumbar (L1–L5), sacral (S1–S5), and coccyx (Cx3–Cx5). The neurological level of injury is the lowest spinal segment with intact motor and sensory function. Injuries are classified as complete or incomplete based on preserved neurological function below the injury level [8]. A complete injury results in total loss of motor and sensory function, while an incomplete injury allows for some function and sensation [3,9,10]. Neurological function alone does not fully define disability; motor and sensory impairments also influence outcomes [11,12]. Standardized assessment tools evaluate SCI severity and recovery potential. The American Spinal Injuries Association (ASIA) impairment scale (AIS) is widely used, classifying injuries from Grade A (complete injury with no function) to Grade E (normal function) based on motor and sensory assessments [13]. ASIA evaluates key muscles and dermatomes, providing a standardized score. The lower extremity motor score (LEMS), a component of ASIA, helps predict walking ability; a score of 30 or higher indicates a good chance of regaining mobility [14]. The complexity of SCI outcomes extends beyond neurological impairment alone and involves multiple dimensions of functioning. Thus, the World Health Organization’s International Classification of Functioning, Disability, and Health (ICF) provides a comprehensive framework to describe health and disability in terms of body function, activity limitation, and social participation [15]. It includes qualifiers for body functions and structures, resulting in a 7-point scale indicating the presence and magnitude (or extent) of an impairment. Integrating the ICF framework allows for a broader interpretation of recovery that goes beyond traditional impairment scales, supporting a more holistic and patient-centered approach to prognosis and rehabilitation in SCI. Research has investigated linking ASIA scores with ICF categories to create standardized outcome measures for spinal cord injury (SCI). For instance, certain ICF codes, such as b730 (muscle power functions) or d450 (walking), can be associated with ASIA motor and sensory scores to measure disability and monitor progress [16]. The ICF’s universal terminology supports interdisciplinary collaboration and international comparisons, while the ASIA classification offers precise, SCI-specific insights [17].

1.3. Emergency and Acute Treatment of SCI

At an accident scene where SCI is suspected, emergency responders prioritize immobilization. A rigid collar is placed around the neck, and the patient is carefully positioned on a backboard [18]. Sedatives may be used to minimize movement, and a breathing tube may be inserted if needed. At a trauma center, interventions include spinal realignment using a brace or mechanical force. Surgery may remove fractured vertebrae, bone fragments, or herniated discs. Early spinal decompression surgery improves functional recovery. SCI often leads to complications requiring treatment [19]. Respiratory issues are common; one-third of SCI patients need temporary or permanent breathing assistance. Injuries at the C1–C4 levels can impair diaphragm function, requiring mechanical ventilation. Pneumonia is a leading cause of death, particularly in patients on ventilators. Circulatory complications (unstable blood pressure, blood clots, and abnormal heart rhythms) necessitate anticoagulants and compression stockings. SCI can also lead to muscle tone changes (spasticity and atrophy). Autonomic dysreflexia, a life-threatening condition in individuals with upper spinal injuries, can cause sudden hypertension, headaches, sweating, and vision disturbances. Positioning the patient upright can help reduce blood pressure. Other complications include pressure ulcers, neurogenic pain, bladder and bowel dysfunction, and sexual health issues, all requiring specialized treatment. SCI also has a psychological impact; many individuals experience depression. Therapy and medication can help manage these mental health challenges [20].

1.4. Rehabilitation and Prognosis in SCI

Rehabilitation programs aim to restore independence and improve quality of life. These programs integrate physical therapies, skill-building activities, and psychological support. A multidisciplinary team tailors treatment to each patient’s needs [21]. Assistive devices (braces, wheelchairs, electronic stimulators, neural prosthetics) can enhance mobility. Predicting recovery from SCI is challenging. Clinical assessments alone often fail to accurately forecast outcomes [22]. Neurophysiological techniques (somatosensory evoked potentials (SSEPs), nerve conduction studies (NCSs), and motor evoked potentials (MEPs)) provide objective measures of neural function [23]. These techniques distinguish nerve damage types and offer prognostic value, particularly in clinical trials. SSEPs are useful in evaluating sensory recovery by assessing the conduction of sensory signals along the spinal cord. Intact or improved SSEP responses suggest better prognosis, while absent or diminished responses indicate severe injury [24]. SSEPs are valuable for assessing injury level, guiding personalized treatment plans, and predicting long-term motor outcomes.

1.5. Electrophysiology and Artificial Intelligence in SCI

Electrophysiology is critical in SCI diagnosis, classification, and prognosis [25]. Among widely used evoked potential (EP) techniques, SSEPs, MEPs, and nerve conduction studies assess neural pathway integrity [26,27]. EPs provide objective insights into neurological dysfunction, even when clinical evaluations are challenging [28,29]. Studies show that SSEPs correlate strongly with SCI outcomes, including walking ability, hand function, and bladder control [30]. Combining clinical assessments (e.g., ASIA scores) with electrophysiological tests enhances diagnostic accuracy and informs treatment strategies. Recent advancements have introduced artificial intelligence (AI) and machine learning (ML) into SCI research. AI leverages large datasets to enhance disease prediction, classification, and treatment planning [31]. ML algorithms use statistical techniques (regression analysis and Bayesian inference) to analyze patient data, improving diagnostic precision. Neural networks have been successfully applied in SCI prognosis, medical imaging, and treatment optimization [6,32]. By integrating AI-driven models with neurophysiological assessments, researchers aim to predict SCI progression and optimize rehabilitation strategies, leading to better patient outcomes.

This study aimed to identify the best predictors of SCI recovery using ML, examining the importance of electrophysiological recordings among other input parameters for ASIA score determination and motor recovery.

2. Materials and Methods

The patient data used are a dataset from a recent investigation of muscle-specific recovery after cervical spinal cord injury [33]. This retrospective analysis of 748 individuals from the European Multicenter Study about Spinal Cord Injury (NCT01571531) showed associations between corticospinal tract (CST) sparing and upper extremity recovery, improving hand muscle strength recovery prediction. The dataset includes 11 input parameters (clinical and full electrophysiological assessment findings—SSEPs, MEPs, and NCVs) and one output: the final SCI recovery result of 123 patients, who were fully recorded for upper extremity recovery. The input parameters are as follows:

• MS: Motor score;
• DST: Distance from the motor level of injury;
• LT: Light touch sensation;
• PP: Pin prick sensation;
• SSEP_Amp diff_uln: Somatosensory evoked potential amplitude difference ulnar nerve;
• Hupp_score_SEP: Somatosensory evoked potential score;
• MEP_Amplitude_abd: Motor evoked potential amplitude recorded from abductor muscle;
• Hupp_score_MEP: Motor evoked potential score;
• F-wave persistance uln: F-wave persistence ulnar nerve;
• Hupp score-NCS: Nerve conduction studies score;
• REC: muscle strength final recovery {recovery class, no recovery class};
• AIS: ASIA score {A, B, C, D, E}.

Some technical/medical tests are defined as follows:

SSEP: Somatosensory evoked potentials (SSEPs) recorded from the central nervous system following electric stimulation of peripheral nerves (upper limb SSEP);

ASIA: The expanded disability status scale (ASIA) quantifies disability in SCI and monitors changes over time;

MEP: Motor evoked potentials recorded from peripheral nerves following magnetic stimulation of the central nervous system;

NCS: Nerve conduction studies using F-wave or compound muscle action potential from the ulnar nerve.

The source database can be found at https://www.synapse.org/Synapse:syn50900348 (accessed on 11 November 2024), created by experts aiming to analyze segmental muscle-specific recovery after cervical spinal cord injury and was extracted from the European Multicenter Study about Spinal Cord Injury as in details described in [33]. Only 123 clinically (MS, DST, LT, PP, and AIS) and “fully electrophysiologically accessed” (SSEP, MEP, and NCS in terms of specific test response details and total scoring) patient records were extracted using their coding information and included in our study. The rest of the records were excluded as partly examined or because important information was missing. The final database used in the present approach can also be found at https://www.synapse.org/Synapse:syn66044903 (accessed on 6 October 2024). Representative statistics of the included patients’ records are presented in Figure 1, where the correlation-based best feature selection revealed that MS, LT, PP, MEP score, and SEP score input parameters are the best predictors for REC and LT and SEP score is the best predictor for AIS [34].

In order to evaluate our hypothesis for patient recovery prediction, we used different ML algorithms and compared their corresponding results based on four metrics commonly used for evaluation of predictions: accuracy, sensitivity, specificity, and precision (abbreviated as Acc, Sen, Spec, and Prec, respectively), defined by the following formulas:

                     Acc = (a + d)/(a + b + c + d)

Sen = a/(a + b)

        Spec = d/(c + d) and

Prec = a/(a + c)

where a is the number of positive (success) cases correctly classified, b is the number of positive cases misclassified as negative (fail), d is the number of negative cases correctly classified, and c is the number of negative cases misclassified as positive.

Accuracy shows how good the system is in predicting correctly (success or failure). Sensitivity shows how good the system is in predicting success, while specificity shows how good the system is in predicting failure. Evaluation results, in terms of ML accuracy scores, are presented in

Table 1. Outcomes derived from statistical accuracy across various clinical and electrophysiological methods and overall assessment predictions.

(a) Ensemble Algorithm (Vote)
Biomarker Accuracy (%)	All Together	SSEPs	MEPs	NCS	Clinical Assessment
Motor Recovery	89.8	85.4	81.3	82.9	75.6
AIS index	84.1	74.9	72.7	74.4	63.4
(b) Randomforest
Biomarker Accuracy (%)	All Together	SSEPs	MEPs	NCS	Clinical Assessment
Motor Recovery	93.1	91.9	86.6	85.4	75.6
AIS index	89.0	80.1	75.2	81.3	66.3
(c) Decision Trees (J48)
Biomarker Accuracy (%)	All Together	SSEPs	MEPs	NCS	Clinical Assessment
Recovery	81.3	79.2	77.9	67.1	73.5
AIS index	71.9	76.0	67.0	67.1	59.8
(d) Neural Networks (Multilayer Perceptron)
Biomarker Accuracy (%)	All Together	SSEPs	MEPs	NCS	Clinical Assessment
Motor Recovery	90.2	82.5	75.2	79.2	71.5
AIS index	78.0	68.7	69.5	65.4	57.7
(e) Bayes (Naive Bayes)
Biomarker Accuracy (%)	All Together	SSEPs	MEPs	NCS	Clinical Assessment
Motor Recovery	76.0	73.9	75.6	71.9	69.5
AIS index	62.6	58.9	58.1	61.4	56.1

and Figure 2 (part of). As in statistics, accuracy is the best metric used to evaluate the performance of a classification model. It measures the proportion of all predictions (both positive and negative) that are correctly classified by the ML model [35].

3. Results

For better interpretation of our hypothesis about the importance of electrophysiology assessment as an SCI patient recovery predictor, different ML algorithms were used. The main approach was ensemble learning, a machine learning technique combining multiple models (regression models, neural networks, and decision trees) to enhance predictive accuracy. This approach integrates several individual models to achieve better results than a single model alone [36]. Research has validated the effectiveness of ensemble learning in machine learning and convolutional neural networks (CNNs). Each machine learning model is influenced by factors including training data, hyper parameters, and other parameters, impacting the total error. Even with the same training algorithm, different models emerge with distinct levels of bias, variance, and irreducible error. Ensemble methods minimize overall error by merging multiple diverse models, preserving each model’s strengths. Studies indicate that ensembles with greater diversity yield more accurate predictions. Ensemble learning effectively mitigates over fitting without significantly increasing model bias. Ensembles of diverse, under-regularized models can outperform individual regularized models [35]. Ensemble techniques can also address challenges related to high-dimensional data. In this study, ensemble algorithms, like vote (integrating decision trees (J48), artificial neural networks (multilayer perceptron), and Bayes (Naïve Bayes)) and random forest, were used. All models were implemented using the Waikato Environment for Knowledge Analysis (WEKA) platform [37]. WEKA is a widely used of machine learning and data analysis free software licensed under the GNU General Public License. It was developed at the University Of Waikato, New Zealand. WEKA version 3.9.0 provides visualization tools and machine learning algorithms for data analysis and predictive modeling. It features graphical user interfaces and data preprocessing functions. WEKA integrates various artificial intelligence techniques and statistical methods, supporting core data mining processes. The platform operates on the principle that data are presented in a structured format, where each instance consists of a defined set of attributes. Many of WEKA’s standard machine learning algorithms generate decision trees for classification tasks.

The first approach was decision trees using the J48 algorithm. Decision trees extract insights and create predictive models. A decision tree is structured like a flowchart, systematically dividing data into branches without information loss. It serves as a hierarchical sorting mechanism, predicting outcomes based on sequential decision-making steps. The tree construction process follows a structured methodology: each node represents a decision point based on a specific parameter, determining the progression to the next branch. This iterative process continues until a leaf node is reached, representing the final predicted outcome (ASIA prediction). To assess the accuracy of the constructed decision tree, the random tree algorithm was used to generate the model based on the dataset.

The second approach was WEKA’s Neural Network (multilayer perceptron with a hidden layer), trained using the error back-propagation algorithm. This intelligent system followed an I–H–O (input–hidden layers–output layers) format [32]. The training process involved gradually increasing the number of neurons in the hidden layer and extending the number of training epochs. With a constant learning rate, a consistent reduction in error per training epoch and improved classification performance were observed. Optimal results were obtained with seven hidden neurons and 15,000 training epochs. Bayesian classifiers, a family of classification algorithms based on Bayes’ Theorem, were also employed. The Naïve Bayes classifier, one of the simplest and most effective, allows for rapid model development and fast prediction. Naïve Bayes is primarily used for classification tasks and is particularly well suited for text classification problems. Its computational efficiency enables swift processing and simplified predictions, even with high-dimensional data. This model estimates the probability that a given instance belongs to a specific class based on a predefined set of features (ASIA score). The dataset was randomly divided (66% records for training, 34% records for testing). Lastly, the random forest algorithm was implemented, building upon the bagging method by incorporating both bagging and feature randomness to create an ensemble of uncorrelated decision trees. Feature randomness ensures low correlation among decision trees by generating a random subset of features. This distinguishes random forest from traditional decision trees, where all potential feature splits are considered; random forest selects only a subset of features for splitting. The random forest algorithm requires setting three hyper parameters before training: node size, the number of trees, and the number of features sampled. Once configured, the classifier can be used for both regression and classification tasks. The model consists of multiple decision trees, each built from a bootstrap sample. Approximately one-third of this sample is reserved as test data (the out-of-bag (OOB) sample), which are later used for validation. Another layer of randomness is introduced through feature bagging, enhancing dataset diversity and reducing correlation among decision trees. The prediction process varies based on the problem type: for regression tasks, individual decision tree outputs are averaged, while for classification tasks, the final class is determined through majority voting. The OOB sample serves as a form of cross-validation, ensuring a reliable prediction outcome.

We tested and present here different algorithm performances to (a) strengthen our hypothesis about electrophysiological assessment and (b) provide different alternatives, using a simpler model, such as Naïve Bayes, or a more interpretable one, such as J48, instead of random forest or ensemble models. This comparison is very helpful, especially considering the balance between performance, interpretability, and computational requirements of each approach [36]. Patient records were divided into training and test sets randomly. The performance of the machine learning models that were developed using the training set was evaluated using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic (ROC) curve (AUC-ROC) in the test set. Part of the results of the models on the testing set is shown in Figure 2 and Table 1. We calculated the prediction accuracy of motor recovery and current ASIA score according to all input parameters. Possible electrophysiology biomarkers (SEPs, MEPs, and all NCSs data) were also compared with clinical evaluation results.

All results reveal the importance of SEPs as predictors of motor recovery and ASIA score in SCI patients. Specific comparisons between SEP, MEP, NCS, and clinical assessment accuracies are detailed here, referencing Table 1 and Figure 3. Electrophysiology assessment, when available, elevates the total accuracy from only clinical investigation to 93.1% (from at most 75.6%) for final motor recovery prediction and to 89% (from at most 66.3%) for ASIA score determination. More data are needed to confirm these results (Table 1).

We suggest that sensory electrophysiology assessment (including SEP and NCV data) is a faster, less expensive, portable, and simpler alternative to other prognostic tests and more effective than clinical assessment methods, such as AIS or ICF scores. By integrating individualized electrophysiological data, we can move toward a precision-medicine approach, tailoring treatment strategies and rehabilitation plans to each patient’s unique neurological profile, ultimately improving functional outcomes.

4. Discussion

Patients with SCI often undergo electrophysiological assessments (MEPs and NCSs) to evaluate neurological deficits and predict recovery potential. Studies have shown significant correlations between electrodiagnostic findings and ASIA scores, with lower scores associated with more severe impairments [30]. Specific electrophysiological parameters enhance the predictive accuracy of ASIA assessments regarding recovery outcomes. Electrophysiological evaluations identify preserved neural pathways, informing rehabilitation strategies and contributing to improved functional recovery [38]. However, relying solely on ASIA scores might overlook subtle nuances of potential recovery, suggesting the need for a more integrated evaluation approach.

Ιn this study, we applied different machine learning algorithms to analyze a dataset consisting of clinical and electrophysiological parameters from individuals with spinal cord injury. Our goal was to evaluate the ability of these models to predict two key outcomes: motor recovery and ASIA impairment classification. Among the models tested, random forest demonstrated the highest accuracy, achieving 93.1% for motor recovery and 89.0% for ASIA classification. These results highlight the potential of combining electrophysiological markers, such as MEP and SEP, with machine learning techniques to improve prognostic assessments in SCI.

While random forest achieved the highest accuracy in both motor recovery (93.1%) and ASIA classification (89.0%), other models such as Naïve Bayes (82.8%, 79.3%) and J48 (86.2%, 82.7%) also performed competitively. These simpler algorithms offer practical advantages: Naïve Bayes is computationally efficient and robust with small datasets, while J48 provides interpretable decision trees that can aid clinicians in understanding and communicating predictive outcomes. In scenarios where model transparency or processing speed is critical—such as real-time clinical decision-making or use in low-resource settings—these models may be preferable despite slightly lower accuracy.

Our research demonstrates that SSEPs have significant prognostic value in predicting motor recovery in SCI patients. Our findings indicate a positive correlation between SSEP latencies and motor recovery, further substantiated by their association with the ASIA impairment scale (AIS). These findings align with existing literature suggesting that SSEPs and MEPs are more sensitive in detecting disease progression than conventional clinical assessments while also offering cost-effective and time-efficient advantages [39]. The integration of machine learning algorithms into SCI prognostic modeling further refines predictive accuracy. By analyzing large-scale datasets, these algorithms can identify complex patterns not immediately discernible to clinicians. Existing rehabilitation protocols could be enhanced with these predictors to create personalized therapy plans. For instance, patients with preserved MEPs may benefit from intensive motor training earlier in the rehabilitation process, while those lacking such responses might be prioritized for compensatory strategies or neuromodulation trials. This approach can support more targeted interventions, optimize resource allocation, and ultimately enhance functional outcomes.

Such advancements optimize the decision-making process and enable evidence-based decisions informed by real-time data analysis. Predictive analytics facilitates early intervention and timely modifications in treatment plans. As these technological innovations progress, they have the potential to transform the management of chronic neurological conditions, paving the way for more personalized, precise, and effective treatment pathways.

The findings of this study are consistent with the two widely used clinical classification systems. ASIA directly corresponds to body function and activity domains, complementing the ICF, which provides a holistic perspective on disability and health. Integrating both frameworks improves the understanding and rehabilitation management of spinal cord injury [17]. The use of objective predictors, such as MEPs and SEPs, can improve early functional assessment, which is vital for guiding interventions that support not only physical recovery but also activity engagement and societal participation. Embedding these predictive models within an ICF-informed rehabilitation strategy could contribute to more comprehensive patient care, integrating functional prognosis with personalized goal-setting and multidisciplinary treatment planning.

Further research is needed to establish EPs as reliable biomarkers for SCI, given their advantages over traditional imaging and biochemical methodologies. EPs are considerably more cost-effective, portable, and straightforward to administer than magnetic resonance imaging (MRI), making them a promising alternative for widespread clinical application [40]. Establishing their efficacy as diagnostic and prognostic tools could enhance accessibility to advanced neurological assessments in diverse healthcare settings. Based on such findings, a decision support system could be developed as an objective alternative to the ASIA scale, using only sensory electrophysiology assessment.

5. Conclusions

This article shows the significant potential of evoked potentials (EPs) as reliable predictors of recovery in SCI patients. By demonstrating a pronounced correlation between EPs and American Spinal Injury Association (ASIA) scores, this research highlights the role of electrophysiological assessments in enhancing clinical prognostication and rehabilitation strategies. The integration of machine learning algorithms reinforces the predictive capabilities of EPs and paves the way for personalized patient care. The implementation of artificial intelligence in analyzing medical data introduces an innovative framework for tailoring treatment plans that can significantly improve patient outcomes. This study’s high accuracy rates for predicting recovery outcomes validate the utility of EPs over traditional assessment methods, emphasizing their cost-effectiveness, portability, and ease of administration. In conclusion, this research advances our understanding of the prognostic value of electrophysiological testing in SCI and suggests a paradigm shift in how healthcare systems can leverage advanced analytics to optimize recovery trajectories for patients. The implications are profound, suggesting a future where enhanced accessibility to reliable diagnostic tools can lead to improved rehabilitation and quality of life for SCI patients. The call for further investigation into sensory electrophysiology assessments is crucial and timely, aiming to solidify these preliminary findings and encourage broader clinical adoption.