An interpretable machine learning model integrating [18F]FDG PET/CT radiomics and clinical features for predicting perforation following chemotherapy in gastrointestinal lymphoma: a multicenter study
A machine learning model combining imaging and blood markers predicted dangerous complications from lymphoma chemotherapy with 85% accuracy.
A multicenter ML model integrating [18F]FDG PET/CT radiomics with clinical features (n=257 GI lymphoma patients) predicted chemotherapy-related GI perforation with external AUC of 0.852 using a logistic regression approach with SHAP interpretability. Key predictors included elevated CRP and T-cell NHL histology, and the combined model outperformed either radiomics or clinical models alone.
What the study was
- Study design
- Retrospective multicenter study with external validation
- Population
- Gastrointestinal lymphoma patients (n=257), two hospitals; PFCGL incidence 12-22% per cohort
- Sample size
- 257
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- European Journal of Nuclear Medicine and Molecular Imaging
Why it surfaced
Clinically relevant multicenter ML model addressing a high-mortality complication (GI perforation) in lymphoma. External validation in 50 patients from a separate institution strengthens generalizability.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.