Сomparative analysis of machine learning models for survival prediction in patients with malignant melanoma.

Authors

Maksym A Chernykh, Olena O Oshyvalova

Published in

Wiadomosci lekarskie (Warsaw, Poland : 1960). Volume 79. Issue 9. Pages 2194-2200.

Abstract

Aim: To perform a comparative analysis of the effectiveness of contemporary machine learning models applied to survival prediction in patients with malignant melanoma and to assess their potential advantages over conventional staging systems as tools for personalised risk stratification.
Materials and Methods: A structured comparative analysis of published prognostic models developed using population-based databases was performed, comparing measures of discrimination (area under the ROC curve [AUC] and concordance index [C-index]), calibration (Brier score), and clinical utility (decision curve analysis). The clinical context was supplemented by data from a retrospective cohort study of patients with cutaneous melanoma.
Results: Ensemble and gradient-boosting algorithms (Random Forest, CatBoost, and XGBoost) consistently outperformed Cox proportional hazards regression models in terms of discrimination, achieving AUC values of approximately 0.76-0.84 for 1-, 3-, and 5-year survival. Ensemble models achieved a C-index of 0.732, compared with 0.685 for Cox regression and 0.537 for AJCC/NCCN criteria; the MIA nomogram demonstrated performance comparable to that of machine learning models (C-index, 0.739). Tumour stage, age, and surgical treatment were consistently identified as the leading prognostic factors. Retrospective cohort data confirmed the prognostic role of Breslow thickness, mitotic activity, and lymph node involvement in disease progression.
Conclusions: Machine learning models provide greater accuracy and more individualised prognostic assessment than conventional approaches and may serve as a supplement to standard staging systems. Their implementation requires prospective validation while preserving the physician's ultimate decision-making authority.

PMID:
43007964
Bibliographic data and abstract were imported from PubMed on 11 Nov 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.