Related Experiment Video
Updated: Jun 16, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Human Versus Machine in Survival Prediction for Metastatic Spinal Cord Compression: A Retrospective Cohort Study
Vaishnavi Sharma1, Rawan Masarwa1, Jonathan Dimitry1
1Centre for Spinal Studies and Surgery, Queens Medical Centre, Nottingham University Hospitals NHS Trust, Nottingham, GBR.
Cureus
|June 15, 2026
Summary
Accurate survival prediction for metastatic spinal cord compression (MSCC) is challenging. AI and surgeon scores improved short-term predictions, but intermediate survival remains difficult to forecast.
Area of Science:
- Oncology
- Medical Informatics
- Spinal Surgery
Background:
- Accurate survival prediction in metastatic spinal cord compression (MSCC) is crucial for treatment planning but remains challenging, especially for intermediate survival durations.
- Current methods rely on clinical judgment and scoring systems, with limited accuracy for specific survival intervals.
Purpose of the Study:
- To compare the accuracy of oncologist judgment, surgeon-calculated Tokuhashi scores, and ChatGPT-assisted predictions in estimating survival outcomes for MSCC patients.
- To identify key predictors of survival in MSCC using machine learning analyses.
Main Methods:
- A retrospective study of 100 MSCC patients from a tertiary spinal oncology center.
- Calculation of surgeon Tokuhashi scores, documentation of oncologist-estimated life expectancy, and generation of ChatGPT-assisted survival predictions.
- Comparison of predictions against actual survival outcomes (<6 months, 6-12 months, >12 months) and machine learning analysis of survival predictors.
Main Results:
- Overall prediction accuracy was highest for ChatGPT Tokuhashi-based predictions (53%), followed by surgeon Tokuhashi scores (49%), oncologist judgment (47%), and ChatGPT literature-based estimates (36%).
- Short-term survival (<6 months) recall was highest with surgeon (70%) and ChatGPT Tokuhashi (68%) methods.
- Intermediate survival (6-12 months) prediction remained difficult across all methods. Oncologists showed better recall for long-term survival (>12 months) at 74%.
- Functional status (Karnofsky score) and patient age were the strongest survival predictors, outperforming tumor type and metastasis burden.
Conclusions:
- Structured prognostic tools and AI-assisted scoring can enhance clinical judgment for short-term survival prediction in MSCC.
- Intermediate-term survival prediction in MSCC represents a significant unmet clinical need.
- Future prognostic strategies should emphasize dynamic functional metrics over static tumor classifications for improved personalized patient care.