Machine learning explainability in nasopharyngeal cancer survival using LIME and SHAP
Problem
Nasopharyngeal cancer patients attain different outcomes and need prognostic stratification into low and high survival groups.
Why it matters
Survival stratification can inform individualized treatment decisions for NPC patients.
What the AI did
A stacked algorithm combining five ML algorithms, with LIME and SHAP explainability, was trained on SEER data and compared against XGBoost for survival stratification.
Results
1094 NPC patients from the SEER database used for training and internal validation; temporal validation performed (performance metrics not stated in source)
Impact
The evidence supports building and validating an explainable ML prognostic system, but not its clinical performance.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Rasheed Omobolaji Alabi, Mohammed Salem Elmusrati, Ilmo V.V. Leivo, Alhadi Almangush, Antti A. Mäkitie |
| Citations | 160 (OpenAlex, at import time) |
| References | 65 |
| Model or method | Stacked ensemble of five ML algorithms with LIME and SHAP |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-023-35795-0 |
| Open access | gold |
| Keywords | nasopharyngeal cancer, survival prediction, explainable AI, LIME, SHAP, machine learning |
Importance rating
real but narrower advance
Prognostic stratification for nasopharyngeal cancer patients could inform treatment decisions, but abstract gives no disease burden or scale; a narrower clinical modeling advance.