Machine learning-based prognostic prediction for cutaneous malignant melanoma patient survival

Area of Science:

• Oncology
• Bioinformatics
• Medical Statistics

Background:

• Cutaneous malignant melanoma (CMM) is a significant health concern requiring early diagnosis and accurate survival prediction.
• Effective prognostic models are crucial for enhancing patient survival rates in CMM.

Purpose of the Study:

• To develop and validate a superior prognostic model for predicting cutaneous malignant melanoma (CMM) survival.
• To compare the performance of the novel DeepSmote model against established machine learning algorithms.

Main Methods:

• A retrospective analysis of 5979 CMM patients from the SEER database (2004-2015) was conducted.
• The SMOTE+DeepSurv (DeepSmote) model was developed and validated using the TCGA dataset, with a 7:3 train-test split.
• Model performance was assessed using metrics including Area Under the Curve (AUC), accuracy, precision, recall, and F1-score, comparing DeepSmote against DeepSurv, XGBoost, LR, SVM, RF, KNN, and DT.

Main Results:

• The DeepSmote model exhibited superior prognostic performance on both SEER and TCGA datasets.
• On the SEER test set, DeepSmote achieved an AUC of 0.96, accuracy of 0.95, and F1-score of 0.95 for 1-year survival prediction.
• Consistent superiority was observed for 3- and 5-year predictions in the external TCGA cohort (AUC: 0.91, accuracy: 0.88, F1-score: 0.87), indicating robustness and generalizability.

Conclusions:

• DeepSmote serves as an accurate and generalizable prognostic tool for CMM survival prediction.
• The model consistently outperformed other evaluated models across multiple datasets and evaluation metrics.
• This advancement holds significant potential for improving clinical decision-making and patient management in CMM care.