Deep Learning Predicts Mutations and Outcomes in Gastrointestinal Stromal Tumors
DL models were trained on WSIs to predict mutations, treatment sensitivity, and recurrence‑free survival (RFS).
In a multi‑center study of 8,398 gastrointestinal stromal tumors (GIST) from 21 institutions, deep learning models trained on whole‑slide images accurately predicted KIT and PDGFRA mutational status (AUC 0.87 and 0.96, respectively), specific sub‑type mutations, and sensitivity to targeted therapies (avapritinib 0.84, imatinib 0.81). The models also forecasted recurrence‑free survival with hazard ratios of 8.44 overall and 4.74 for patients on adjuvant therapy, matching conventional pathology risk scores. These results demonstrate that histology‑based deep learning can reliably identify key genomic alterations, predict therapeutic response, and stratify prognosis in GIST.
Accurate, non‑invasive prediction of driver mutations, drug sensitivity and recurrence risk from routine histology would streamline clinical decision‑making in GIST.
Evidence level: Feltételezés / hírjelzés. Nincs önálló tudományos bizonyíték.
Related signals
- Disproportionality analysis of adverse events associated with tyrosine kinase inhibitors in gastrointestinal stromal tumors in the FDA adverse event reporting system
- Review of Genomic Testing and SDH Deficiency in Gastrointestinal Stromal Tumors: Getting to the GIST
- Molecular Tailored Therapeutic Options for Advanced Gastrointestinal Stromal Tumors (GISTs): Current Practice and Future Perspectives
- Phase II study of neoadjuvant imatinib in large gastrointestinal stromal tumours of the stomach
- UK clinical practice guidelines for the management of gastrointestinal stromal tumours (GIST)