Next Article in Journal
“Ultrasound-Assisted High-Fluidics Phacoaspiration”: An Efficient and Safe Technique for Nucleus Quadrant Workup Through 2.2 mm Mini- and 1.6 mm Micro-Incisions Following Manual or Femtosecond Laser Sectorial Lens Fragmentation
Previous Article in Journal
Investigation of Optic Nerve Sheath Diameter, Intraocular Pressure, and Dry Eye in Patients with Borderline Personality Disorder: The Role of Childhood Trauma
Previous Article in Special Issue
Applications and Performance of Artificial Intelligence in Spinal Metastasis Imaging: A Systematic Review
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
JCM
Volume 14
Issue 16
10.3390/jcm14165885
jcm-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Systematic Review

Artificial Intelligence Models for Predicting Outcomes in Spinal Metastasis: A Systematic Review and Meta-Analysis

by
Vivek Sanker
1,*,
Prachi Dawer
2,
Alexander Thaller
3,
Zhikai Li
4,
Philip Heesen
5,
Srinath Hariharan
1,
Emil O. R. Nordin
1,
Maria Jose Cavagnaro
1,
John Ratliff
1 and
Atman Desai
1
1
Department of Neurosurgery, Stanford University, Palo Alto, CA 94305, USA
2
Department of Neurosurgery, University College of Medical Sciences, New Delhi 110095, India
3
Department of Neurosurgery, Medical University of Graz, 8010 Graz, ST, Austria
4
Department of Clinical Neurosciences, Addenbrooke’s Hospital, Cambridge CB2 0QQ, UK
5
Faculty of Medicine, University of Zurich, 8006 Zurich, Switzerland
*
Author to whom correspondence should be addressed.
J. Clin. Med. 2025, 14(16), 5885; https://doi.org/10.3390/jcm14165885
Submission received: 10 June 2025 / Revised: 5 August 2025 / Accepted: 13 August 2025 / Published: 20 August 2025
(This article belongs to the Special Issue Recent Advances in Spine Tumor Diagnosis and Treatment)
Browse Figures
Versions Notes

Abstract

Background: Spinal metastases can cause significant impairment of neurological function and quality of life. Hence, personalized clinical decision-making based on prognosis and likely outcome is desirable. The effectiveness of AI in predicting complications and treatment outcomes for patients with spinal metastases is assessed. Methods: A thorough search was carried out through the PubMed, Scopus, Web of Science, Embase, and Cochrane databases up until 27 January 2025. Included were studies that used AI-based models to predict outcomes for adult patients with spinal metastases. Three reviewers independently extracted the data, and screening was conducted in accordance with PRISMA principles. AUC results were pooled using a random-effects model, and the PROBAST program was used to evaluate the study’s quality. Results: Included were 47 articles totaling 25,790 patients. For training, internal validation, and external validation, the weighted average AUCs were 0.762, 0.876, and 0.810, respectively. The Skeletal Oncology Research Group machine learning algorithms (SORG-MLAs) were the ones externally validated the most, continuously producing AUCs > 0.84 for 90-day and 1-year mortality. Models based on radiomics showed promise in preoperative planning, especially for outcomes of radiation and concealed blood loss. Most research concentrated on breast, lung, and prostate malignancies, which limited its applicability to less common tumors. Conclusions: AI models have shown reasonable accuracy in predicting mortality, ambulatory status, blood loss, and surgical complications in patients with spinal metastases. Wider implementation necessitates additional validation, data standardization, and ethical and regulatory framework evaluation. Future work should concentrate on creating multimodal, hybrid models and assessing their practical applications.

1. Introduction

The spine is one of the most common sites of metastasis, after the lung and the liver. In patients with systemic cancer, approximately more than half develop spinal metastases, and approximately 10% are symptomatic [1]. Surgery can significantly improve quality of life in selected patients [2], and overall treatments have advanced in recent years to enhance overall clinical results and survival [3]. However, the likelihood of good outcomes must always be weighed against the risks of complications and the economic costs in each individual case [4].
Artificial intelligence (AI) is emerging as a potentially powerful tool to enhance clinical decision-making through analysis of large datasets to predict individual patient outcomes and risks, through machine learning and deep learning algorithms.
This systematic review’s main goal was to assess the state of artificial intelligence (AI) models created to forecast outcomes and problems for patients who have spinal metastases. The degree to which these models included explainability and interpretability—two crucial components for clinical confidence and the practical application of AI tools—was specifically examined in addition to summarizing performance measures. The following definition of the primary clinical outcomes was made to guarantee uniformity between studies:
  • Survival: At the longest or most precise follow-up available, it is reported as either overall survival (OS) or progression-free survival (PFS). PFS is the period of time until disease progression or death, whereas OS is the period of time from diagnosis or the start of therapy to death from any cause.
  • Ambulatory status: Usually classified as either ambulatory or non-ambulatory, this refers to the patient’s capacity to walk on their own or with the use of assistive technology.
  • Complications: Contains any unfavorable events that occur during or after surgery, such as bleeding, infection, thrombosis, or neurological decline. Standard classification systems (e.g., Clavien–Dindo) were used to stratify these by severity whenever possible.

2. Materials and Methods

2.1. Ethical Review

Ethical review and approval were waived for this study due to it being a systematic review of previously published data that did not involve human participants or the collection of new data.

2.2. Search Strategy

We searched PubMed, Scopus, Web of Science Advance, Cochrane, and Embase (Ovid) databases to identify relevant studies, using a search query with specific keywords like ‘spine metastases’, ‘artificial intelligence’, ‘machine learning’, ‘deep learning’, and ‘outcomes’ (Supplementary Table S1). The population under consideration included adults with spinal metastases. The objective was to identify studies reporting the use of AI/deep learning (DL) models in predicting treatment and outcome prediction in spinal metastases [5].
Irrelevant articles, such as studies unrelated to spinal metastases and those purely investigating primary spinal tumors, were excluded. Animal studies, reviews, and non-original research articles were also excluded from our analysis to ensure the inclusion of primary research data relevant to our objective. The electronic search ranged from the period’s earliest available date up to 27 January 2025 [5].

2.3. Screening of Studies

Each study’s title and abstract were screened for relevance before proceeding to full-text screening, which was independently assessed by two reviewers (PD and VS). Any discrepancies were addressed through consultation with a third reviewer (SH). This review adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines but was not registered on the PROSPERO international prospective register of systematic reviews (Figure 1).

2.4. Data Extraction

Three independent authors (PD, VS, and AT) extracted relevant data from the included studies. The data collected included study design, participant demographics, and the number of participants with respective outcomes and complications. Discrepancies in data extraction were resolved through consensus [5].

2.5. Data Analysis

Relevant variables were extracted from each of the included articles, such as the primary tumor type, cohort size, and prediction model performance matrices: area under the receiver operating characteristic curve (AUC) and type of validation (internal or external validation). The weighted average of the AUC was calculated. All statistical analyses were conducted using Excel, R Statistical Software, version 4.3.1, and Python, version 3.13.3.
We conducted a random-effects meta-analysis using restricted maximum likelihood (REML) estimation to account for both within-study and between-study variability. This approach assumes that the true effect size may vary across studies due to underlying differences in study populations, methodologies, or settings. REML was used to estimate the between-study variance (τ2), providing an unbiased and efficient estimate of heterogeneity. Pooled effect estimates were calculated as weighted averages of the individual study effects, with weights derived from both the within-study variance and the estimated between-study variance. The Standardized Mean Difference (SMD) along with the 95% confidence interval (CI) was used to compare continuous performance metrics like AUC between studies after having a pooled estimate.

2.6. Quality Assessment

The quality assessment was performed using the PROBAST (Prediction model Risk of Bias Assessment Tool (Supplementary Figures S1 and S2).
PROBAST is designed for assessing the risk of bias and applicability in studies that develop, validate, or update predictive models. It is a structured tool that assesses four domains:
Participants: evaluating whether the data sources or patient samples used for training and testing are appropriate and representative of the clinical population. Predictors: ensuring that input data or predictors are well defined and appropriately measured. Outcome: ensuring that the outcomes (e.g., model predictions, decisions) are clearly defined and relevant to clinical scenarios. Analysis: evaluating whether the model performance metrics, training/validation processes, and statistical analysis methods are robust and unbiased.
In the Participants Section, 32 studies were flagged as having low risk of bias (68%), whilst 15 studies were flagged as having unclear risk/some concerns (32%). In the Predictors Section, 23 studies were flagged as having low risk of bias (49%), 16 studies flagged as having unclear risk/some concerns (34%), and 8 studies flagged as having high risk (17%). In the Outcome Section, 18 studies were flagged as having low risk of bias (38%), 16 as having unclear risk/some concerns (34%), and 13 as having high risk (28%). In the Analysis Section, 3 studies were flagged as having low risk of bias (6%), 15 studies flagged as having unclear risk/some concerns (32%), and 29 flagged as having high risk (62%).

3. Results

This review encompasses 47 studies published between 2016 and 2025, including a total of 26,038 patients with a median of 269 patients per study, ranging from 30 to 2786 patients (Table 1).
Among the 47 studies, the three most common primary tumor types were breast cancer, lung cancer and prostate cancer, reported in 33 (70.2%), 32 (68.1%), and 23 (48.9%) studies, respectively (Table 2 and Figure 2). In contrast, neuroendocrine tumors, bladder cancer and esophageal cancer were the least common, each reported in three (6.4%), three (6.4%), and four (8.5%) studies, respectively (Table 2).
Five of the forty-seven studies reported AUC values for the established models during training of the model (Table 3), eighteen for internal validation (Table 4), and fourteen for external validation (Table 5).
The weighted average AUC value among the five studies that reported AUC values and corresponding 95% confidence intervals for the training of the established models is 0.762 (95% CI: 0.704–0.717). Wherever 95% confidence intervals were not reported, the weighted average of the reported 95% confidence intervals was used.
The weighted average AUC value among the 18 studies that reported AUC values and corresponding 95% confidence intervals for internal validation of the established models is 0.876 (95% CI: 0.871–0.881). Wherever 95% confidence intervals were not reported, the weighted average of the reported 95% confidence intervals was used.
The weighted average AUC value among the eight studies that reported AUC values and corresponding 95% confidence intervals for external validation of the established models is 0.810 (95% CI: 0.803–0.816).

Meta-Analysis of the SORG-MLA Model

The pooled AUC of the SORG-MLA model for 90-day survival was 0.79 (95% CI: 0.75–0.82) and for 1-year survival 0.80 (95% CI: 0.75–0.85). The prediction intervals ranged from 0.65 to 0.88 and from 0.62 to 0.91, respectively. Figure 3 and Figure 4 are forest plots of the 90-day survival and 1-year survival.

4. Discussion

The integration of artificial intelligence (AI) into clinical oncology is revolutionizing the care of patients with spinal metastases. Here, data were combined from 47 studies with a pool of >25,000 patients with a range of models presented for a breadth of clinical applications, from predicting surgical mortality and complications to estimating occult blood loss and assessing perioperative functional status. The heterogeneity of use-cases highlights the versatility of AI for addressing different facets of managing spinal metastases. Notably, prediction models demonstrated strong discriminative performance with pooled AUC values of 0.762, 0.876, and 0.810 for training, internal, and external validation, respectively, supporting the reasonable accuracy and generalizability of these models across multiple datasets (Table 3, Table 4 and Table 5). Additionally, the predominance of certain primary tumor types, including breast, lung, and prostate, within the included studies reflects their real-world contribution to spinal metastases. While this focus ensures relevance to a large patient population, it also raises questions about generalizability of these models to rarer malignancies which might be underrepresented, including neuroendocrine or esophageal sources. There is a need for more balanced datasets and the inclusion of between- and within-subgroup analyses to account for tumor-type-specific variability in outcomes and treatment response.
Reproducibility remains a fundamental challenge in radiomics-based models due to variability in image acquisition protocols, reconstruction parameters, and scanner types across institutions. These inconsistencies can substantially alter extracted radiomic features, thereby affecting model performance when applied outside the development cohort [37]. Our review noted that only a minority of studies implemented harmonization techniques such as image resampling, intensity normalization, or ComBat-based feature adjustment, which aim to mitigate site-specific variability. The lack of standardization across studies limits generalizability and may contribute to overfitting or unstable performance in external validation cohorts [37]. Future research should prioritize standardized radiomics pipelines and transparent reporting of acquisition parameters to enable reliable replication and multi-center deployment of radiomics-based predictive tools.
Crucially, survival prediction emerged as the most extensively studied application of AI in spinal metastasis, with several model algorithms, especially those developed by the Skeletal Oncology Research Group (SORG), subject to both internal and international external validation. The consistent performance of these models across different populations and healthcare infrastructures supports their utility as decision-support tools [23,24,30,33]. For instance, models predicting 90-day and 1-year mortality yielded AUCs approaching or exceeding 0.84 in many studies, making a strong case for these tools to be used in preoperative planning or when evaluating eligibility for aggressive intervention [21,23,24,35,36]. Despite this, all studies that externally evaluated the SORG-MLA were retrospective in their design, thereby limiting the real-world significance of the pooled AUCs. Therefore, the realization of real-world impact studies is encouraged to assess clinical endpoints such as changes in treatment decisions and improved survival.
Although AI models’ prognostic abilities in spinal metastases have shown promise, the discipline is still struggling with the crucial problems of interpretability and explainability. The safe and moral incorporation of AI into healthcare decision-making depends on these factors. The fact that many of the evaluated research rely on opaque, sophisticated algorithms, including deep neural networks and ensemble-based models like XGBoost, random forest, CatBoost, or Gaussian processes, without providing clear reasoning routes, is a major drawback [12,13,15,17,18,20,21]. A model must be interpretable in order to promote patient acceptance, clinician trust, and regulatory compliance in the therapeutic setting, where decisions have significant consequences [3,10].
Because it is difficult to understand how input features affect predictions, there is rising worry that “black box” models might be dangerous for patient care [10,15,17]. However, only a small percentage of the papers we examined used explainable AI (XAI) methods to help physicians and stakeholders understand the model’s logic, such as SHAP (Shapley Additive Explanations), LIME (Local Interpretable Model-Agnostic Explanations), or decision-path visualizations. For example, there was little investigation into the methods by which high-performing models created with random forest [10,22], support vector machines [13], and deep learning [19] arrived at their predictions.
Adoption in multidisciplinary decision-making settings [5], when physicians are held responsible for defending treatment decisions, may be constrained by this lack of openness. It will be crucial to include model interpretability tools (such as attention heatmaps and feature significance plots) into upcoming development workflows. AI in spine cancer cannot be meaningfully implemented without clinician-centered design, explainable model architecture, and cross-validation across a variety of datasets [3,25].
Adoption and clinical implementation of AI models is largely dependent on the ease of use and availability of such models. We identified six studies [12,15,21,23,24,27] that deployed a user-friendly platform, i.e., web-based applications. Of these studies, only two web-based applications [21,24] were working trouble-free at the time of the creation of this manuscript. The two functioning applications represent the SORG-MLA at different time points, which have been integrated into one website [24]. Such availability builds the foundation on which clinical uptake and regular usage is dependent. Therefore, more models should be made readily available by deploying them via user-friendly platforms, such as websites, mobile apps etc.

5. Limitations

We have now added a specific Limitations Section to offer a thorough and open evaluation of the limitations of this research. Regarding data preparation, outcome definitions, validation methods, and reporting standards, the included studies exhibit methodological variability. To reduce direct cross-comparability of model performance, for example, some research used split-sample validation or no validation at all, while others used k-fold cross-validation [14,15,21].
Second, there is significant worry about demographic and regional bias across datasets. The bulk of training cohorts come from middle- to high-income nations like China, Taiwan, and the United States [21,30,31,34], which can restrict the applicability of AI tools to underrepresented groups or environments with limited resources. International external validation studies are available [31,33,34], but few of them specifically address equality, cultural adaptability, or the diversity of healthcare systems when using AI. Further external validation should thus be carried out to ensure applicability and global generalizability of existing models. Ethnic or racial biases are addressed in some studies; however, the homogenous nature of many validation cohorts does not allow for adequate distinctions of model performance between different ethnic or racial groups to be made. Future models should therefore be trained on a heterogenous pool of patients to account for cultural, ethnic, racial, and socioeconomic differences.
Third, while some models showed good discrimination metrics (AUCs > 0.85), we were unable to fully evaluate the clinical value of these tools since calibration metrics and decision-curve analysis were seldom published [23,25,32]. It is noteworthy that the majority of models are still in the research realm and have not yet been included into real-time clinical decision processes or electronic health records (EHRs2). The limited number of studies that documented attempts to create web-based apps or clinician-facing tools [15,38] limited its translational preparedness.
Lastly, the examined literature still lacks sufficient attention to ethical and practical concerns, such as patient privacy, data completeness, interpretability, and openness. For instance, in several studies, missing variables (such as albumin or lymphocyte counts) deteriorated model performance [27], but few models included flexible topologies or robust imputation to deal with such uncertainty. Future multicenter, prospective, inclusive studies that assess the practical performance, equity, and stakeholder acceptance of AI tools in spinal cancer are desperately needed, as these data highlight.

6. Conclusions

AI has the potential to advance spinal metastasis care in the domains of outcome prediction and risk stratification and thus enhance precision medicine and streamline clinical workflow to improve patient outcomes. If rigorously validated and ethically deployed, AI has the potential not just to predict outcomes but to transform spinal oncology into a more personalized, equitable, and data-driven discipline. As AI continues to evolve within spinal oncology, its success will depend not only on predictive accuracy but also on transparency, interpretability, and ethical deployment. Future models must prioritize explainability to foster clinician trust, support informed consent, and ensure equitable care across diverse patient populations.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/jcm14165885/s1, Supplementary Figure S1: Risk of bias. Supplementary Figure S2: Applicability. Supplementary Table S1: Search queries across databases. Supplementary Table S2: Summary of the studies analyzed. References [39,40,41,42,43,44,45,46,47,48,49,50,51,52,53] are cited in the supplementary materials.

Author Contributions

Conceptualization, V.S.; Methodology, V.S.; Formal Analysis, V.S. and A.T.; Investigation, V.S., P.D., Z.L. and P.H.; Data Curation, V.S., P.D.; Writing—Original Draft Preparation, V.S.; Writing—Review and Editing, V.S., P.D., A.T., Z.L., P.H., S.H., E.O.R.N., M.J.C., J.R. and A.D.; Supervision, J.R. and A.D.; Project Administration, vs. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Ethical review and approval were waived for this study due it being a systematic review of previously published data and did not involve human participants or the collection of new data.

Data Availability Statement

All datasets used for this study are presented in the manuscript and its Supplementary Materials.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Wewel, J.T.; O’Toole, J.E. Epidemiology of spinal cord and column tumors. Neuro-Oncol. Pract. 2020, 7 (Suppl. S1), i5–i9. [Google Scholar] [CrossRef]
  2. Conti, A.; Acker, G.; Kluge, A.; Loebel, F.; Kreimeier, A.; Budach, V.; Vajkoczy, P.; Ghetti, I.; Germano’, A.F.; Senger, C.; et al. Decision Making in Patients with Metastatic Spine. The Role of Minimally Invasive Treatment Modalities. Front. Oncol. 2019, 9, 915. [Google Scholar] [CrossRef]
  3. Dixon, D.; Sattar, H.; Moros, N.; Kesireddy, S.R.; Ahsan, H.; Lakkimsetti, M.; Fatima, M.; Doshi, D.; Sadhu, K.; Junaid Hassan, M. Unveiling the Influence of AI Predictive Analytics on Patient Outcomes: A Comprehensive Narrative Review. Cureus 2024, 16, e59954. [Google Scholar] [CrossRef]
  4. Barton, L.B.; Arant, K.R.; Blucher, J.A.; Sarno, D.L.; Redmond, K.J.; Balboni, T.A.; Colman, M.; Goodwin, C.R.; Laufer, I.; Placide, R.; et al. Clinician Experiences in Treatment Decision-Making for Patients with Spinal Metastases. J. Bone Jt. Surg. 2021, 103, e1. [Google Scholar] [CrossRef] [PubMed]
  5. Sanker, V.; Sanikommu, S.; Thaller, A.; Li, Z.; Heesen, P.; Hariharan, S.; Nordin, E.O.R.; Cavagnaro, M.J.; Ratliff, J.; Desai, A. Artificial Intelligence for Non-Invasive Prediction of Molecular Signatures in Spinal Metastases: A Systematic Review. Bioengineering 2025, 12, 791. [Google Scholar] [CrossRef]
  6. Zhao, W.; Qin, S.; Wang, Q.; Chen, Y.; Liu, K.; Xin, P.; Lang, N. Assessment of Hidden Blood Loss in Spinal Metastasis Surgery: A Comprehensive Approach with MRI-Based Radiomics Models. J. Magn. Reson. Imaging 2024, 59, 2023–2032. [Google Scholar] [CrossRef] [PubMed]
  7. Bakhsheshian, J.; Shahrestani, S.; Buser, Z.; Hah, R.; Hsieh, P.C.; Liu, J.C.; Wang, J.C. The performance of frailty in predictive modeling of short-term outcomes in the surgical management of metastatic tumors to the spine. Spine J. 2022, 22, 605–615. [Google Scholar] [CrossRef]
  8. Massaad, E.; Bridge, C.P.; Kiapour, A.; Fourman, M.S.; Duvall, J.B.; Connolly, I.D.; Hadzipasic, M.; Shankar, G.M.; Andriole, K.P.; Rosenthal, M.; et al. Evaluating frailty, mortality, and complications associated with metastatic spine tumor surgery using machine learning–derived body composition analysis. J. Neurosurg. Spine 2022, 37, 263–273. [Google Scholar] [CrossRef]
  9. Shi, Y.J.; Zhu, H.T.; Li, X.T.; Zhang, X.Y.; Wei, Y.Y.; Yan, S.; Sun, Y.S. Radiomics analysis based on multiple parameters MR imaging in the spine: Predicting treatment response of osteolytic bone metastases to chemotherapy in breast cancer patients. Magn. Reson. Imaging 2022, 92, 10–18. [Google Scholar] [CrossRef]
  10. Massaad, E.; Williams, N.; Hadzipasic, M.; Patel, S.S.; Fourman, M.S.; Kiapour, A.; Schoenfeld, A.J.; Shankar, G.M.; Shin, J.H. Performance assessment of the metastatic spinal tumor frailty index using machine learning algorithms: Limitations and future directions. Neurosurg. Focus 2021, 50, E5. [Google Scholar] [CrossRef]
  11. Santipas, B.; Suvithayasiri, S.; Trathitephun, W.; Wilartratsami, S.; Luksanapruksa, P. Developmental and Validation of Machine Learning Model for Prediction Complication After Cervical Spine Metastases Surgery. Clin. Spine Surgery: A Spine Publ. 2024, 38, E81–E88. [Google Scholar] [CrossRef]
  12. Cui, Y.; Shi, X.; Qin, Y.; Wang, Q.; Cao, X.; Che, X.; Pan, Y.; Wang, B.; Lei, M.; Liu, Y. Establishment and validation of an interactive artificial intelligence platform to predict postoperative ambulatory status for patients with metastatic spinal disease: A multicenter analysis. Int. J. Surg. 2024, 110, 2738–2756. [Google Scholar] [CrossRef]
  13. Santipas, B.; Chanajit, A.; Wilartratsami, S.; Ittichaiwong, P.; Veerakanjana, K.; Luksanapruksa, P. Development of Machine Learning Algorithms for Predicting Preoperative and Postoperative venous Thromboembolism in Patients Undergoing Surgery for Spinal Metastasis. Siriraj Med. J. 2024, 76, 381–388. [Google Scholar] [CrossRef]
  14. Santipas, B.; Veerakanjana, K.; Ittichaiwong, P.; Chavalparit, P.; Wilartratsami, S.; Luksanapruksa, P. Development and internal validation of machine-learning models for predicting survival in patients who underwent surgery for spinal metastases. Asian Spine J. 2024, 18, 325–335. [Google Scholar] [CrossRef]
  15. Shi, X.; Cui, Y.; Wang, S.; Pan, Y.; Wang, B.; Lei, M. Development and validation of a web-based artificial intelligence prediction model to assess massive intraoperative blood loss for metastatic spinal disease using machine learning techniques. Spine J. 2024, 24, 146–160. [Google Scholar] [CrossRef] [PubMed]
  16. Chavalparit, P.; Wilartratsami, S.; Santipas, B.; Ittichaiwong, P.; Veerakanjana, K.; Luksanapruksa, P. Development of Machine-Learning Models to Predict Ambulation Outcomes Following Spinal Metastasis Surgery. Asian Spine J. 2023, 17, 1013–1023. [Google Scholar] [CrossRef] [PubMed]
  17. Chen, Y.; Qin, S.; Zhao, W.; Wang, Q.; Liu, K.; Xin, P.; Yuan, H.; Zhuang, H.; Lang, N. MRI feature-based radiomics models to predict treatment outcome after stereotactic body radiotherapy for spinal metastases. Insights Into Imaging 2023, 14, 169. [Google Scholar] [CrossRef] [PubMed]
  18. Gao, L.; Cao, Y.; Cao, X.; Shi, X.; Lei, M.; Su, X.; Liu, Y. Machine learning-based algorithms to predict severe psychological distress among cancer patients with spinal metastatic disease. Spine J. 2023, 23, 1255–1269. [Google Scholar] [CrossRef]
  19. Hallinan, J.T.P.D.; Zhu, L.; Zhang, W.; Kuah, T.; Lim, D.S.W.; Low, X.Z.; Cheng, A.J.L.; Eide, S.E.; Ong, H.Y.; Muhamat Nor, F.E.; et al. Deep Learning Model for Grading Metastatic Epidural Spinal Cord Compression on Staging CT. Cancers 2022, 14, 3219. [Google Scholar] [CrossRef]
  20. Jabehdar Maralani, P.; Chen, H.; Moazen, B.; Mojtahed Zadeh, M.; Salehi, F.; Chan, A.; Zeng, L.K.; Abugharib, A.; Tseng, C.L.; Husain, Z.; et al. Proposing a quantitative MRI-based linear measurement framework for response assessment following stereotactic body radiation therapy in patients with spinal metastasis. J. Neurooncol. 2022, 160, 265–272. [Google Scholar] [CrossRef]
  21. Karhade, A.V.; Fenn, B.; Groot, O.Q.; Shah, A.A.; Yen, H.K.; Bilsky, M.H.; Hu, M.H.; Laufer, I.; Park, D.Y.; Sciubba, D.M.; et al. Development and external validation of predictive algorithms for six-week mortality in spinal metastasis using 4304 patients from five institutions. Spine J. 2022, 22, 2033–2041. [Google Scholar] [CrossRef]
  22. Gui, C.; Chen, X.; Sheikh, K.; Mathews, L.; Lo, S.L.; Lee, J.; Khan, M.A.; Sciubba, D.M.; Redmond, K.J. Radiomic modeling to predict risk of vertebral compression fracture after stereotactic body radiation therapy for spinal metastases. J. Neurosurg. Spine 2022, 36, 294–302. [Google Scholar] [CrossRef]
  23. Karhade, A.V.; Thio, Q.C.B.S.; Ogink, P.T.; Shah, A.A.; Bono, C.M.; Oh, K.S.; Saylor, P.J.; Schoenfeld, A.J.; Shin, J.H.; Harris, M.B.; et al. Development of Machine Learning Algorithms for Prediction of 30-Day Mortality After Surgery for Spinal Metastasis. Neurosurgery 2019, 85, E83–E91. [Google Scholar] [CrossRef]
  24. Karhade, A.V.; Thio, Q.C.B.S.; Ogink, P.T.; Bono, C.M.; Ferrone, M.L.; Oh, K.S.; Saylor, P.J.; Schoenfeld, A.J.; Shin, J.H.; Harris, M.B.; et al. Predicting 90-Day and 1-Year Mortality in Spinal Metastatic Disease: Development and Internal Validation. Neurosurgery 2019, 85, E671–E681. [Google Scholar] [CrossRef]
  25. Paulino Pereira, N.R.; Janssen, S.J.; van Dijk, E.; Harris, M.B.; Hornicek, F.J.; Ferrone, M.L.; Schwab, J.H. Development of a Prognostic Survival Algorithm for Patients with Metastatic Spine Disease. J. Bone Jt. Surg. 2016, 98, 1767–1776. [Google Scholar] [CrossRef] [PubMed]
  26. Fenn, B.P.; Karhade, A.V.; Groot, O.Q.; Collins, A.K.; Balboni, T.A.; Oh, K.S.; Ferrone, M.L.; Schwab, J.H. Survival in Patients with Spinal Metastatic Disease Treated Nonoperatively with Radiotherapy. Clin. Spine Surgery A Spine Publ. 2024, 37, E290–E296. [Google Scholar] [CrossRef] [PubMed]
  27. Huang, C.C.; Peng, K.P.; Hsieh, H.C.; Groot, O.Q.; Yen, H.K.; Tsai, C.C.; Karhade, A.V.; Lin, Y.P.; Kao, Y.T.; Yang, J.J.; et al. Does the Presence of Missing Data Affect the Performance of the SORG Machine-learning Algorithm for Patients with Spinal Metastasis? Development of an Internet Application Algorithm. Clin. Orthop. Relat. Res. 2024, 482, 143–157. [Google Scholar] [CrossRef] [PubMed]
  28. Pan, Y.T.; Lin, Y.P.; Yen, H.K.; Yen, H.H.; Huang, C.C.; Hsieh, H.C.; Janssen, S.; Hu, M.H.; Lin, W.H.; Groot, O.Q. Are Current Survival Prediction Tools Useful When Treating Subsequent Skeletal-related Events From Bone Metastases? Clin. Orthop. Relat. Res. 2024, 482, 1710–1721. [Google Scholar] [CrossRef] [PubMed]
  29. Li, Z.; Guo, L.; Guo, B.; Zhang, P.; Wang, J.; Wang, X.; Yao, W. Evaluation of different scoring systems for spinal metastases based on a Chinese cohort. Cancer Med. 2023, 12, 4125–4136. [Google Scholar] [CrossRef]
  30. Su, C.C.; Lin, Y.P.; Yen, H.K.; Pan, Y.T.; Zijlstra, H.; Verlaan, J.J.; Schwab, J.H.; Lai, C.Y.; Hu, M.H.; Yang, S.H.; et al. A Machine Learning Algorithm for Predicting 6-Week Survival in Spinal Metastasis: An External Validation Study Using 2768 Taiwanese Patients. J. Am. Acad. Orthop. Surg. 2023, 31, e645–e656. [Google Scholar] [CrossRef]
  31. Zhong, G.; Cheng, S.; Zhou, M.; Xie, J.; Xu, Z.; Lai, H.; Yan, Y.; Xie, Z.; Zhou, J.; Xie, X.; et al. External validation of the SORG machine learning algorithms for predicting 90-day and 1-year survival of patients with lung cancer-derived spine metastases: A recent bi-center cohort from China. Spine J. 2023, 23, 731–738. [Google Scholar] [CrossRef]
  32. Yen, H.K.; Hu, M.H.; Zijlstra, H.; Groot, O.Q.; Hsieh, H.C.; Yang, J.J.; Karhade, A.V.; Chen, P.C.; Chen, Y.H.; Huang, P.H.; et al. Prognostic significance of lab data and performance comparison by validating survival prediction models for patients with spinal metastases after radiotherapy. Radiother. Oncol. 2022, 175, 159–166. [Google Scholar] [CrossRef] [PubMed]
  33. Shah, A.A.; Karhade, A.V.; Park, H.Y.; Sheppard, W.L.; Macyszyn, L.J.; Everson, R.G.; Shamie, A.N.; Park, D.Y.; Schwab, J.H.; Hornicek, F.J. Updated external validation of the SORG machine learning algorithms for prediction of ninety-day and one-year mortality after surgery for spinal metastasis. Spine J. 2021, 21, 1679–1686. [Google Scholar] [CrossRef] [PubMed]
  34. Yang, J.J.; Chen, C.W.; Fourman, M.S.; Bongers, M.E.R.; Karhade, A.V.; Groot, O.Q.; Lin, W.H.; Yen, H.K.; Huang, P.H.; Yang, S.H.; et al. International external validation of the SORG machine learning algorithms for predicting 90-day and one-year survival of patients with spine metastases using a Taiwanese cohort. Spine J. 2021, 21, 1670–1678. [Google Scholar] [CrossRef]
  35. Bongers, M.E.R.; Karhade, A.V.; Villavieja, J.; Groot, O.Q.; Bilsky, M.H.; Laufer, I.; Schwab, J.H. Does the SORG algorithm generalize to a contemporary cohort of patients with spinal metastases on external validation? Spine J. 2020, 20, 1646–1652. [Google Scholar] [CrossRef]
  36. Karhade, A.V.; Ahmed, A.K.; Pennington, Z.; Chara, A.; Schilling, A.; Thio, Q.C.B.S.; Ogink, P.T.; Sciubba, D.M.; Schwab, J.H. External validation of the SORG 90-day and 1-year machine learning algorithms for survival in spinal metastatic disease. Spine J. 2020, 20, 14–21. [Google Scholar] [CrossRef]
  37. Lee, S.B.; Hong, Y.; Cho, Y.J.; Jeong, D.; Lee, J.; Choi, J.W.; Hwang, J.Y.; Lee, S.; Choi, Y.H.; Cheon, J.E. Enhancing Radiomics Reproducibility: Deep Learning-Based Harmonization of Abdominal Computed Tomography (CT) Images. Bioengineering 2024, 11, 1212. [Google Scholar] [CrossRef]
  38. He, X.; Jiao, Y.Q.; Yang, X.G.; Hu, Y.C. A Novel Prediction Tool for Overall Survival of Patients Living with Spinal Metastatic Disease. World Neurosurg. 2020, 144, e824–e836. [Google Scholar] [CrossRef]
  39. Kehayias, C.E.; Bontempi, D.; Quirk, S.; Friesen, S.; Bredfeldt, J.; Kosak, T.; Kearney, M.; Tishler, R.; Pashtan, I.; Huynh, M.A.; et al. A prospectively deployed deep learning-enabled automated quality assurance tool for oncological palliative spine radiation therapy. Lancet Digit. Health 2024, 7, e13–e22. [Google Scholar] [CrossRef]
  40. Li, J.; Zhang, J.; Zhang, X.; Lun, D.; Li, R.; Ma, R.; Hu, Y. Quantile regression-based prediction of intraoperative blood loss in patients with spinal metastases: Model development and validation. Eur. Spine J. 2023, 32, 2479–2492. [Google Scholar] [CrossRef]
  41. Mezei, T.; Horváth, A.; Nagy, Z.; Czigléczki, G.; Banczerowski, P.; Báskay, J.; Pollner, P. A Novel Prognostication System for Spinal Metastasis Patients Based on Network Science and Correlation Analysis. Clin. Oncol. 2022, 35, e20–e29. [Google Scholar] [CrossRef]
  42. Fourman, M.S.; Siraj, L.; Duvall, J.; Ramsey, D.C.; De La Garza Ramos, R.; Hadzipasic, M.; Connolly, I.; Williamson, T.; Shankar, G.M.; Schoenfeld, A.; et al. Can We Use Artificial Intelligence Cluster Analysis to Identify Patients with Metastatic Breast Cancer to the Spine at Highest Risk of Postoperative Adverse Events? World Neurosurg. 2023, 174, e26–e34. [Google Scholar] [CrossRef]
  43. Li, Z.; Huang, L.; Guo, B.; Zhang, P.; Wang, J.; Wang, X.; Yao, W. The predictive ability of routinely collected laboratory markers for surgically treated spinal metastases: A retrospective single institution study. BMC Cancer. MC Cancer 2022, 22, 1231. [Google Scholar] [CrossRef]
  44. Hu, M.H.; Yen, H.K.; Chen, I.H.; Wu, C.H.; Chen, C.W.; Yang, J.J.; Wang, Z.Y.; Yen, M.H.; Yang, S.H.; Lin, W.H. Decreased psoas muscle area is a prognosticator for 90-day and 1-year survival in patients undergoing surgical treatment for spinal metastasis. Clin. Nutr. 2022, 41, 620–629. [Google Scholar] [CrossRef]
  45. Walker, A.; Bassale, S.; Shukla, R.; Dai Kubicky, C. A Prognostic Index for Predicting Survival of Patients Undergoing Radiation Therapy for Spine Metastasis Using Recursive Partitioning Analysis. J. Palliat. Med. 2022, 25, 21–27. [Google Scholar] [CrossRef]
  46. Khalid, S.I.; Massaad, E.; Kiapour, A.; Bridge, C.P.; Rigney, G.; Burrows, A.; Shim, J.; De la Garza Ramos, R.; Tobert, D.G.; Schoenfeld, A.J.; et al. Machine learning–based detection of sarcopenic obesity and association with adverse outcomes in patients undergoing surgical treatment for spinal metastases. J. Neurosurg. Spine 2024, 40, 291–300. [Google Scholar] [CrossRef] [PubMed]
  47. Rigney, G.H.; Massaad, E.; Kiapour, A.; Razak, S.S.; Duvall, J.B.; Burrows, A.; Khalid, S.I.; De La Garza Ramos, R.; Tobert, D.G.; Williamson, T.; et al. Implication of nutritional status for adverse outcomes after surgery for metastatic spine tumors. J. Neurosurg. Spine 2023, 39, 557–567. [Google Scholar] [CrossRef] [PubMed]
  48. Hallinan, J.T.P.D.; Zhu, L.; Zhang, W.; Lim, D.S.W.; Baskar, S.; Low, X.Z.; Yeong, K.Y.; Teo, E.C.; Kumarakulasinghe, N.B.; Yap, Q.V.; et al. Deep Learning Model for Classifying Metastatic Epidural Spinal Cord Compression on MRI. Front. Oncol. 2022, 12, 849447. [Google Scholar] [CrossRef] [PubMed]
  49. Arends, S.R.S.; Savenije, M.H.F.; Eppinga, W.S.C.; van der Velden, J.M.; van den Berg, C.A.T.; Verhoeff, J.J.C. Clinical utility of convolutional neural networks for treatment planning in radiotherapy for spinal metastases. Phys. Imaging Radiat. Oncol. 2022, 21, 42–47. [Google Scholar] [CrossRef]
  50. Rogé, M.; Henni, A.H.; Neggaz, Y.A.; Mallet, R.; Hanzen, C.; Dubray, B.; Colard, E.; Gensanne, D.; Thureau, S. Evaluation of a Dedicated Software “ElementsTM Spine SRS, Brainlab®” for Target Volume Definition in the Treatment of Spinal Bone Metastases with Stereotactic Body Radiotherapy. Front Oncol. 2022, 12, 827195. [Google Scholar] [CrossRef]
  51. Kowalchuk, R.O.; Waters, M.R.; Richardson, K.M.; Spencer, K.; Larner, J.M.; McAllister, W.H.; Sheehan, J.P.; Kersh, C.R. Stereotactic body radiation therapy for spinal metastases: A novel local control stratification by spinal region. J. Neurosurg. Spine 2021, 34, 267–276. [Google Scholar] [CrossRef]
  52. Korpics, M.C.; Polley, M.Y.; Bhave, S.R.; Redler, G.; Pitroda, S.P.; Luke, J.J.; Chmura, S.J. A Validated T Cell Radiomics Score Is Associated with Clinical Outcomes Following Multisite SBRT and Pembrolizumab. Int. J. Radiat. Oncol. Biol. Phys. 2020, 108, 189–195. [Google Scholar] [CrossRef]
  53. Cui, Y.; Pan, Y.; Lei, M.; Mi, C.; Wang, B.; Shi, X. The First Algorithm Calculating Cement Injection Volumes in Patients with Spine Metastases Treated with Percutaneous Vertebroplasty. Ther. Clin. Risk Manag. 2020, 16, 417–4288. [Google Scholar] [CrossRef]
Jcm 14 05885 g001
Figure 1. A PRISMA flow diagram is presented to illustrate the screening of studies.
Figure 1. A PRISMA flow diagram is presented to illustrate the screening of studies.
Jcm 14 05885 g001
Jcm 14 05885 g002
Figure 2. The percentage of how many studies feature specific primary tumor types is presented.
Figure 2. The percentage of how many studies feature specific primary tumor types is presented.
Jcm 14 05885 g002
Jcm 14 05885 g003
Figure 3. A Forest plot of the meta-analysis of the AUC for the SORG-MLA model for 90-day survival is shown. The AUC and 95% Confidence Intervals of each independent model validation are depicted. The diamond represents the pooled AUC along with a 95% Confidence Interval. A prediction interval is presented [26,27,28,29,31,32,33,34,35,36].
Figure 3. A Forest plot of the meta-analysis of the AUC for the SORG-MLA model for 90-day survival is shown. The AUC and 95% Confidence Intervals of each independent model validation are depicted. The diamond represents the pooled AUC along with a 95% Confidence Interval. A prediction interval is presented [26,27,28,29,31,32,33,34,35,36].
Jcm 14 05885 g003
Jcm 14 05885 g004
Figure 4. Forest plot of the meta-analysis of the AUC for the SORG-MLA model for 1-year survival. The AUC and 95% Confidence Intervals of each independent model validation are depicted. The diamond represents the pooled AUC along with a 95% Confidence Interval. A prediction interval is presented [26,27,28,29,31,32,33,34,35,36].
Figure 4. Forest plot of the meta-analysis of the AUC for the SORG-MLA model for 1-year survival. The AUC and 95% Confidence Intervals of each independent model validation are depicted. The diamond represents the pooled AUC along with a 95% Confidence Interval. A prediction interval is presented [26,27,28,29,31,32,33,34,35,36].
Jcm 14 05885 g004
Table 1. An overview of the studies analyzed is presented.
Table 1. An overview of the studies analyzed is presented.
Year of publishing (range)2016–2025
Total number of patients25,790
Median number of patients268
Number of patients per study (range)30–2786
Table 2. The frequency of primary tumor types is shown. Studies analyzing multiple primary tumor types have been included in all applicable categories.
Table 2. The frequency of primary tumor types is shown. Studies analyzing multiple primary tumor types have been included in all applicable categories.
Primary Tumor TypeNumber of Studies (Percentage)
Breast cancer33/47 (70.2%)
Lung cancer32/47 (68.1%)
Prostate cancer23/47 (48.9%)
Esophageal cancer4/47 (8.5%)
Bladder cancer3/47 (6.4%)
Neuroendocrine tumors3/47 (6.4%)
Table 3. AUC (training of each study’s best model).
Table 3. AUC (training of each study’s best model).
StudyOutput/PredictionBest-Performing ModelAUC95% CI
Zhao et al. (2024) [6]Hidden blood loss in spinal metastasis surgeryMRI-Based Radiomics0.784-
Bakhsheshian et al. (2022) [7]MortalityMachine Learning Model using ECI 1 and Frailty0.788-
Medical complications0.723-
Massaad et al. (2022) [8]1-year mortalityMachine Learning Model using Body Composition and NESMS 20.730.67–0.78
Shi et al. (2022) [9]Response of osteolytic metastases to chemotherapyRadiomics (T2WI + ADCall)0.9080.86–0.96
Massaad et al. (2021) [10]Postoperative complicationsRandom Forest to develop MSTFI 30.620.56–0.68
The AUC and corresponding 95% confidence interval of studies reporting AUC values for the established radiomics models for the training of the model is presented. 1 = Elixhauser Comorbidity Index, 2 = New England Spinal Metastasis Score, 3 = Metastatic Spinal Tumor Frailty Index.
Table 4. AUC (internal validation of each study’s best model).
Table 4. AUC (internal validation of each study’s best model).
StudyOutput/PredictionBest-Performing ModelAUC95% CI
Santipas et al. (2025) [11]Complications after cervical spine metastases surgeryGradient Boosting0.939 10.873 2--
Cui et al. (2024) [12]Postoperative ambulatory statusEnsemble Machine Learning combining LR 6, eXGBM 7, SVM 8, RF 9, NN 10 and DT 110.911 10.854–0.968 1
Santipas et al. (2024) [13]30-day preoperative VTE 5Gradient Boosted Trees0.77 1-
90-day preoperative VTE 5Support Vector Machine0.72 1-
30-day postoperative VTE 5Gradient Boosted Trees0.71 1-
90-day postoperative VTE 5Support Vector Machine0.68 1-
Santipas et al. (2024) [14]90-day survivalCatBoost0.750 10.758 2--
180-day survivalXGBoost0.726 10.744 2--
365-day survivalXGBoost0.731 10.693 2-
Shi et al. (2024) [15]Massive intraoperative blood lossXGBoosting machine (XGBM; Machine Learning)0.857 20.827–0.877 2
Zhao et al. (2024) [6]Hidden blood lossMRI-Based Radiomics0.744 20.576–0.914 2
Chavalparit et al. (2023) [16]90-day postoperative ambulatory statusDecision Tree0.941 1-
180-day postoperative ambulatory statusExtreme Gradient Boosting0.852 1-
Chen et al. (2023) [17]Treatment outcome after stereotactic body RT 12Gaussian Processes0.828 1-
Gao et al. (2023) [18]Severe psychological distressGradient Boosting Machine (Machine Learning)0.865 20.788–0.941 2
Hallinan et al. (2022) [19]Grading metastatic epidural spinal cord compression (Bilsky grading (normal/low versus high)Separated Window Learning (Max fusion model)0.971 20.961–0.981 2
Grading metastatic epidural spinal cord compression (Bilsky grading (normal versus low/high)Separated Window Learning (Spine-window)0.924 20.910–0.938 2
Jabehdar Maralani et al. (2022) [20]Response following stereotactic body RTDecision Tree0.923 10.959 2--
Karhade et al. (2022) [21]6-week mortalityElastic-net penalized logistic regression0.85 10.84 20.84–0.86 10.80–0.88 2
Shi et al. (2022) [9]Response of osteolytic metastases to chemotherapyRadiomics (FST2WI + ADCall)0.873 20.78–0.96 2
Gui et al. (2021) [22]Risk of vertebral compression fracture after stereotactic body RT 12Random Forest0.878 10.832–0.924 1
Massaad et al. (2021) [10]Postoperative complicationsRandom Forest to develop MSTFI 30.69 40.66–0.73 4
Karhade et al. (2019) [23]30-day mortalityBayes Point Machine (Machine Learning)0.786 10.782 2--
Karhade et al. (2019) [24]90-day mortalityStochastic gradient boosting0.83 10.83 20.81–0.85 1-
1-year mortality0.85 10.89 20.83–0.87 1-
Paulino-Pereira et al. (2016) [25]30-day survivalBoosting AlgorithmNomogram0.91 10.75 20.86–0.95 10.60–0.89 2
90-day survival0.86 10.73 20.83–0.90 10.63–0.83 2
1-year survival0.84 10.75 20.80–0.87 10.67–0.84 2
The AUC and corresponding 95% confidence interval of studies reporting AUC values for the established radiomics models for internal validation of the model is presented. 1 = k-fold cross validation, 2 = split sample internal validation, 3 = Metastatic Spinal Tumor Frailty Index, 4 = Internal Bootstrap Validation, 5 = venous thromboembolism, 6 = Logistic Regression, 7 = Extreme Gradient Boosting Machine, 8 = Support Vector Machine, 9 = Random Forest, 10 = Neural Network, 11 = Decision Tree, 12 = Radiotherapy.
Table 5. AUC (external validation of each study’s best model).
Table 5. AUC (external validation of each study’s best model).
StudyOutput/PredictionBest Performing ModelAUC95% CI
Cui et al. (2024) [12]Postoperative ambulatory statusEnsemble Machine Learning combining LR 2, eXGBM 3, SVM 4, RF 5, NN 6 and DT 70.8730.809–0.936
0.9240.890–0.959
Fenn et al. (2024) [26]90-day mortalityMachine Learning Algorithm (SORG-MLA)0.850.83–0.87
1-year mortality0.870.85–0.89
Huang et al. (2024) [27]6-week survivalMachine Learning Algorithm (SORG-MLA)0.840.78–0.89
90-day survival0.840.79–0.90
1-year survival0.770.73–0.80
Pan et al. (2024) [28]42-day survivalMachine Learning Algorithm (SORG-MLA)0.690.63–0.74
90-day survival0.720.66–0.77
1-year survival0.700.61–0.78
Shi et al. (2024) [15]Massive intraoperative blood loss for spinal metastasesXGBoosting machine (XGBM; Machine Learning)0.8090.778–0.860
Li et al. (2023) [29]90-day survivalMachine Learning Algorithm (SORG-MLA)0.7430.666–0.817
180-day survivalMachine Learning Algorithm (Revised Katagiri)0.7610.696–0.826
1-year survivalMachine Learning Algorithm (SORG-MLA)0.7870.730–0.838
2-year survivalMachine Learning Algorithm (Revised Katagiri)0.7790.747–0.811
Su et al. (2023) [30]6-week survival after RT 1 onlyMachine Learning Algorithm (SORG-MLA)0.770.74–0.79
6-week survival after surgery0.840.79–0.90
Zhong et al. (2023) [31]90-day mortalityMachine Learning Algorithm (SORG-MLA)0.7140.589–0.839
1-year mortality0.8320.758–0.906
Karhade et al. (2022) [21]6-week mortalityElastic-net penalized logistic regression0.820.78–0.85
Yen et al. (2022) [32]90-day survivalMachine Learning Algorithm (SORG-MLA)0.780.76–0.80
1-year survival0.760.74–0.78
Shah et al. (2021) [33]90-day mortalityMachine Learning Algorithm (SORG-MLA)0.840.79–0.89
1-year mortality0.900.86–0.93
Yang et al. (2021) [34]90-day mortalityMachine Learning Algorithm (SORG-MLA)0.730.67–0.78
1-year mortality0.740.69–0.79
Bongers et al. (2020) [35]90-day mortalityMachine Learning Algorithm (SORG-MLA)0.810.74–0.87
1-year mortality0.840.77–0.89
Karhade et al. (2020) [36]90-day mortalityMachine Learning Algorithm (SORG-MLA)0.810.70–0.89
1-year mortality0.780.67–0.87
The AUC and corresponding 95% confidence interval of studies reporting AUC values for the established radiomics models for external validation of the model is presented. 1 = Radiation Therapy 2 = Logistic Regression, 3 = Extreme Gradient Boosting Machine, 4 = Support Vector Machine, 5 = Random Forest, 6 = Neural Network, 7 = Decision Tree.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Sanker, V.; Dawer, P.; Thaller, A.; Li, Z.; Heesen, P.; Hariharan, S.; Nordin, E.O.R.; Cavagnaro, M.J.; Ratliff, J.; Desai, A. Artificial Intelligence Models for Predicting Outcomes in Spinal Metastasis: A Systematic Review and Meta-Analysis. J. Clin. Med. 2025, 14, 5885. https://doi.org/10.3390/jcm14165885

AMA Style

Sanker V, Dawer P, Thaller A, Li Z, Heesen P, Hariharan S, Nordin EOR, Cavagnaro MJ, Ratliff J, Desai A. Artificial Intelligence Models for Predicting Outcomes in Spinal Metastasis: A Systematic Review and Meta-Analysis. Journal of Clinical Medicine. 2025; 14(16):5885. https://doi.org/10.3390/jcm14165885

Chicago/Turabian Style

Sanker, Vivek, Prachi Dawer, Alexander Thaller, Zhikai Li, Philip Heesen, Srinath Hariharan, Emil O. R. Nordin, Maria Jose Cavagnaro, John Ratliff, and Atman Desai. 2025. "Artificial Intelligence Models for Predicting Outcomes in Spinal Metastasis: A Systematic Review and Meta-Analysis" Journal of Clinical Medicine 14, no. 16: 5885. https://doi.org/10.3390/jcm14165885

APA Style

Sanker, V., Dawer, P., Thaller, A., Li, Z., Heesen, P., Hariharan, S., Nordin, E. O. R., Cavagnaro, M. J., Ratliff, J., & Desai, A. (2025). Artificial Intelligence Models for Predicting Outcomes in Spinal Metastasis: A Systematic Review and Meta-Analysis. Journal of Clinical Medicine, 14(16), 5885. https://doi.org/10.3390/jcm14165885

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop