Regarding the diagnostic and prognostic value of other imaging techniques in the field of SCI, diffusion imaging is a promising technique that can offer more detailed imaging of the injury compared to conventional magnetic resonance imaging (Middendorp et al. 2011). Additionally, the neurological examination according to the International Standards for Neurological Classification of Spinal Cord Injury has become the cornerstone for assessing the severity and level of injury. As for treatment, it has been noted that despite promising progress in basic research for spinal cord restoration, there is currently no effective treatment leading to significant neurological or functional recovery after SCI. Nevertheless, significant advances have been made in the care of patients with SCI during the 21st century, including the prevention of complications and the introduction of specialized care by pioneers in SCI rehabilitation, such as Dr. Donald Munro and Sir Ludwig Guttmann, which has led to increased survival rates in the SCI population.
However, in the current, there have been tremendous strides in the field of assessment, diagnosis, and prognosis of SCIs. Compared to established predictive steps, the last five years have witnessed real progress, with significant contributions from technology and Artificial Intelligence. An excellent example is the Spinal Cord Injury Risk Score - SCIRS, which was developed to estimate the mortality risk for patients who have suffered thoracic SCIs compared to the Injury Severity Score - ISS, a general trauma measurement (Fallah et al. 2022). The analysis showed that age, American Spinal Injury Association Impairment Scale (AIS) classification, neurological level of injury, spinal column morphology, and associated injuries were significant predictors of early mortality after tSCI. The ability to predict mortality using a simple, fast, and reliable assessment tool upon a patient's admission to the healthcare setting would greatly assist in making timely clinical decisions and improving the outcome of incidents.
Thus, current prognostic tools, such as the Injury Severity Score, which predicts mortality after trauma, do not adequately consider the unique characteristics of traumatic SCIs. Fallah et al., in a study conducted in 2021, used machine learning techniques on patient data to develop the Spinal Cord Injury Risk Score (SCIRS) that can predict mortality based on age, neurological level and type of injury, SCI, and Abbreviated Injury Scale scores in comparison to the performance of the Injury Severity Score (ISS), a measure used for predicting mortality after general trauma. The results showed that SCIRS can predict in-hospital mortality and one-year mortality after SCI with higher accuracy than ISS. SCIRS can be used in research to reduce bias in parameter estimation and can help in adjusting coefficients when developing models (Fallah et al. 2022).
It is worth mentioning that developing a mortality prediction model poses challenges due to the complex interactions of factors contributing to patient outcomes. Models based on generalized linear models have been used in the past to develop predictive tools in various clinical studies (Kirshblum et al. 2011; Frankel et al. 1969). Nevertheless, despite the advantage of simplicity with directly available and interpretable parameters, these models may not capture the possible interactions and complex behavior of variables often present in biological conditions. Acute traumatic SCI involves primary and secondary injury mechanisms (Witiw and Fehlings 2015). The primary mechanism is related to the initial traumatic damage caused by the catastrophic impact, and this damage is irreversible. The secondary mechanisms, which start a few minutes after the initial injury, include processes such as ischemia of the spinal cord, excitotoxicity, ionic dysregulation, and oxidative stress caused by free radicals (Eckert and Martin 2017). SCI is characterized by different forms of injury, where the exploration of pathology and clinical diagnosis, therapeutic strategies, animal models that have allowed for a better understanding of injury mechanisms, and finally, the role of new diagnostic and prognostic tools, such as miRNA, could improve the management of this traumatology (Pinchi et al. 2019; Yong et al. 2019).
Predictive Value of Biomarkers
Due to a recent review by Schading et al., published in July (2021), developments in the search for clinically significant biomarkers in SCI are presented. SCI is a complex and heterogeneous condition that can lead to a wide range of functional impairments. The current clinical evaluations of SCI are limited in their ability to predict outcomes and guide therapeutic decisions. As a result, there is an increasing need for more sophisticated assessments and the development of biomarkers that can complement current clinical measurements (Schading et al. 2021). Further studies have identified several potential biomarkers for SCI, including advanced neuroimaging techniques and molecular biomarkers. These biomarkers promise to predict outcomes, monitor disease progression, and guide therapeutic decisions. However, further validation is required before these biomarkers can be applied in clinical practice. The term "biomarkers" in the field of SCI refers to advanced neuroimaging and molecular biomarkers that are sensitive to the detection of this condition (Rodrigues and Moura-Neto 2018). To elaborate, these biomarkers range from advanced neuroimaging techniques to neurophysiological indicators and molecular biomarkers that identify concentrations of various proteins in the blood and cerebrospinal fluid (Leister et al. 2020).
Clinical assessment with standardized neurological examination is the gold standard for assessing the severity of the injury and predicting functional outcomes in SCIs as revised by ASIA and ISCoS International Standards Committee (2019). Thus, these models can be improved by including advanced diagnostic methods, indicating that a multiparametric approach—including neuroimaging and cerebrospinal fluid/ blood biomarkers—improves the accuracy of predicting individual recovery trajectories. In other words, these biomarkers can complement current clinical evaluations by providing additional information that can enhance the accuracy of predicting individual recovery trajectories (Schading et al. 2021). Biomarkers, in general, can be categorized into structural and inflammatory factors, as well as indicators measured in routine blood analyses. Structural biomarkers are mostly cell-type-specific proteins from neural tissue that leak into the cerebrospinal fluid and blood after injury. These tissue-specific proteins are produced by different cells, such as neurons or glial cells. Following SCI, changes in the concentrations of several of these proteins have been observed in both blood and cerebrospinal fluid. Inflammatory biomarkers include cytokines, chemokines, and other factors related to the immune system and are produced in response to injury. Routine blood analysis indicators include the number of white blood cells, C-reactive protein, and the erythrocyte sedimentation rate (Schading et al. 2021).
Furthermore, neurophysiological techniques such as measuring nerve conduction, motor evoked potentials, and SSEPs provide objective measures of neural integrity and allow differentiation between demyelination and axonal damage (Li et al. 2021; Abdelkader et al. 2019). Their value as independent tools for stratifying patients with SCI into subgroups and their prognostic utility have already been demonstrated and validated several years ago. Thus, some have questioned whether these electrophysiological parameters could add valuable information to improve the prediction of functional outcomes. In summary, recent developments in identifying reliable biomarkers for traumatic SCI and improving prognostic models are promising. Clinical evaluation with standardized neurological examination remains the mainstay for assessing the severity of the injury and predicting functional outcomes. However, these models can be improved by including advanced diagnostic methods, indicating that a multiparametric approach—including neuroimaging and blood indicators—enhances the accuracy of predicting individual recovery trajectories (Freund et al. 2019). A plethora of studies exploring the exact potential of this approach with multivariable models capable of accommodating multimodal data to demonstrate the usefulness of these advanced biomarker combinations is deemed necessary.
According to Schading et al., the inclusion of electrophysiological multiparametric parameters in the prediction model leads to better accuracy in forecasting. The research suggests that the assessment of neurological function and prognostic accuracy in patients with SCI can be improved by adding neurophysiological methods to standardized clinical evaluation (Schading et al. 2021). Regarding electrophysiological outcome measures in clinical trials for SCI, in 64 articles that met eligibility criteria, assessing 877 individuals with SCI who received various interventions and 324 individuals with and without SCI who served as controls, five types of clinical trial study designs were identified, with hybrid designs that included both controls and crossover interventions (Korupolu et al. 2019). The use of the Delphi method to develop consensus on standardized guidelines for collecting and reporting electrophysiological results in SCI clinical trials stood out. The Delphi method is a process of achieving group consensus by providing experts with questionnaires and group responses before each subsequent round. Examples of electrophysiological measurements used in SCI clinical trials include cortical somatosensory evoked potentials, motor evoked potentials and spinal reflexes. The results are based on the significance of reporting parameters such as amplitude, latency, and optimal stimulus intensity for obtaining motor-evoked potentials (Korupolu et al. 2019).
Electrophysiological measures have many benefits in SCI clinical trials (Sand et al. 2013). They are largely objective, independent of patient cooperation, and unbiased, as the results do not depend on subjective patient responses. Electrophysiological measures can also provide information about the neurophysiology of SCI, which can guide future therapies that may subsequently achieve clinically significant results. Additionally, electrophysiological measures can be used in combination with conventional clinical outcome measures to provide a more comprehensive assessment of treatment efficacy. Therefore, the literature suggests that future studies should use standardized protocols for data collection and analysis, such as the Delphi method, and report parameters such as amplitude, latency, and optimal stimulus intensity for proper EP acquisition.
Evidence of Somatosensory Evoked Potentials in Spinal Cord Injuries
According to the literature, the majority of the studies regarding the prognostic value of SSEPs in SCIs show a positive correlation with patient recovery (Fustes et al. 2021). It has been found that the contribution of SSEPs is particularly significant in predicting recovery after SCI, either independently or in conjunction with other examinations (ASIA), or as a specialized tool occasionally used for objective differentiation of SCIs, aiding in distinguishing incomplete from complete injuries, especially in patients who are comatose or uncooperative (Li et al. 2021). It has been demonstrated that changes in SSEPs can reflect changes in gross motor function and fine motor function after mild SCI and that changes in the EP amplitude may also reflect changes in fine motor function after severe SCI (Li et al. 2021).
Patients with acute SCI in spinal shock are more sensitive to the assessment of relative damage to the peripheral motor pathways, i.e., the motor neurons and nerve roots (Singh et al. 2020). Recordings from electromyography, nerve conduction studies, and reflex reproduction allow for predicting increased muscle tone or muscle atrophy in comparison to clinical examination. The evaluation of damage to the autonomic nervous system after SCI with clinical examination is limited. Conversely, recordings of Sympathetic Skin Response - SSR can provide information about the extent and level of damage to the sympathetic nervous system related to autonomic dysfunction (Kumru et al. 2009). Responses recorded from the scalp hair are absent in complete cervical SCIs, while incomplete injuries produce various abnormalities in SSEPs. SSEPs can help localize the sensory level in cases of injury and moreover aid in determining the prognosis for functional outcomes. Furthermore, early recording of an SSEP from the tibia has been associated with favorable functional and neurological status and outcome after SCI (Chawla et al. 2019). Therefore, electrophysiological recordings, as complementary to the clinical examination, are useful for designing and selecting the appropriate therapeutic approach for the rehabilitation program. Additionally, they allow for predicting functional outcomes and providing objective assessment regarding spinal and peripheral pathway recovery.
Somatosensory Evoked Potentials in the Diagnosis of Spinal Cord Injuries
As established above, SSEPs are neurophysiological tests used in the assessment and prognosis of SCI (Mauromatis 1996). SSEPs measure electrical activity as a response to sensory stimulation and can provide information about the integrity and function of sensory pathways. Therefore, they are used directly for the diagnosis of the injury and indirectly as a prognostic factor (Kakulas 2004). Regarding the prediction of sensory recovery, SSEPs can assist in evaluating the potential for sensory recovery after spinal cord injury. By measuring the conduction of sensory signals along the spinal cord, SSEPs can indicate the presence or absence of sensory transmission from the level of the injury and below. If SSEPs show intact or improved sensory responses, this indicates a better prognosis for sensory recovery (Zeiler and Koenig 2013).
Determining the level of injury, which can be identified through neurological examination and imaging methods, can be confirmed using SSEPs, providing more detailed data on the damage that the spinal cord suffered. By stimulating specific nerves or dermatomes and recording the resulting responses, SSEPs can determine the segmental level of sensory dysfunction and correlate it with the corresponding level of spinal cord injury (Nardone et al. 2015). Additionally, assessing the severity of the injury is of utmost importance to be performed as early as possible, to make decisions about limiting secondary damage in the affected area. A particular feature of SSEPs is that they can provide necessary information even if the patient is in a comatose state. Decreased or absent SSEP responses indicate significant damage to the sensory pathways and may indicate a more severe injury.
Nevertheless, it is worth noting that SSEPs are a part of a comprehensive evaluation process, and prognosis is determined based on multiple factors, including clinical examination, imaging, and other neurophysiological tests.
Somatosensory Evoked Potentials as Therapy for Spinal Cord Injuries
SSEPs are not typically used as a direct therapy for SCI. SSEPs are primarily used as diagnostic tools to assess the integrity of sensory pathways and provide information about the level and severity of the injury. They are used during the diagnostic phase to help healthcare professionals understand the extent of the injury and guide the development of an appropriate treatment plan. SSEPs, along with other diagnostic tests and clinical evaluations, provide valuable information for determining the course of treatment, such as surgical intervention, rehabilitation interventions, or medical management (Schwab and Bartholdi 2006; Fehlings and Vaccaro 2019).
However, ongoing research efforts are exploring potential therapeutic interventions for SCI, including regenerative medicine, electrical stimulation, and other emerging approaches. While SSEPs can be used as part of the assessment process for these experimental therapies, they are not the therapy itself (Ahuja and Fehlings 2016).
Machine Learning in the Diagnosis of Spinal Cord Injuries
Personalized medicine is a model of a much better medical approach where interventions are based on individual patient characteristics rather than guidelines. As epidemiological datasets continue to grow in size and complexity, robust methods such as statistical machine learning and Artificial Intelligence (AI) become necessary for the interpretation and development of prognostic models from underlying data. Through such analysis, machine learning can facilitate personalized medicine through its accurate predictions (Khan et al. 2020). Additionally, other AI tools, such as natural language processing and computer vision, can play a crucial role in customizing care for patients with SCIs.
Traumatic SCI and degenerative changes in the spine that cause compression of the spinal cord or nerve roots are the two main categories of diseases treated by spine surgeons. SCI results in catastrophic physical, vocational, and psychosocial consequences for almost 180,000 patients worldwide each year. The damage suffered by the spinal cord in the context of the injury, combined with the limited ability of nervous tissue to regenerate, can occasionally lead to irreversible neurological consequences (Khan et al. 2020). Based on the data accumulated so far, the use of predictive algorithms such as machine learning could provide significant preoperative information to both doctors and patients regarding the outcome and the likelihood of adverse events of surgical treatment. As a result, instead of being the outcome of a general analysis, treatments can be personalized, taking into account individual characteristics, relevant factors affecting outcomes, and comorbidities.
It is now a fact that the field of AI has profoundly influenced many industries, including healthcare (Dietz et al. 2022). Its ability to recognize patterns and self-correct to improve over time mimics human cognitive function but on a much larger scale. Machine learning (ML), a subset of AI, ranges in complexity from classical ML to unsupervised ML to deep learning, where Natural Language Processing and Computer Vision are possible. AI-based tools have been developed for segmenting spinal structures, obtaining basic measurements of the spine, and even detecting pathologies such as tumors or degeneration. AI algorithms could be used to guide clinical management by aiding treatment selection, predicting outcomes for individual patients, and even powering neuroprosthetic devices after SCI. While the use of AI has its pitfalls and must be adopted with caution, its future use is promising in the field of spine surgery and medicine as a whole (Katsuura et al. 2021; Katsos et al. 2023). A diagram of knowledge discovery by data collectors in new guidelines is following (Fig. 2).