Related Experiment Videos
Nonlinear discriminant analysis and prognostic factor classification in node-negative primary breast cancer using
J M Le Goff1, L Lavayssière, J Rouëssé
1AVENTI, Marseille, France.
Anticancer Research
|August 6, 2000
Summary
Non-linear kernel discriminant analysis accurately predicted breast cancer recurrence. Combining urokinase-type plasminogen activator (uPA) and clinical tumor size improved prediction accuracy, aiding clinical decision-making for node-negative breast cancer patients.
Area of Science:
- Oncology
- Biostatistics
- Machine Learning
Background:
- Predicting outcomes for 134 axillary node-negative primary breast cancer patients.
- Focus on patients not receiving adjuvant therapy.
- Utilized a non-censored database for analysis.
Purpose of the Study:
- To predict the outcome of primary breast cancer patients.
- To identify key prognostic factors for cancer recurrence.
- To develop a robust predictive model for clinical application.
Main Methods:
- Employed non-linear kernel discriminant analysis (KDA).
- Used probabilistic neural networks (PNN) and leave-one-out cross-validation for relapse probability estimation.
- Applied a stepwise method to select prognostic factors, including tumor characteristics and protein levels, evaluated by ROC indicator.
Main Results:
- The best one-dimensional model used urokinase-type plasminogen activator (uPA) (ROC indicator = 0.75).
- A two-factor model with uPA and clinical tumor size (T) achieved the highest discrimination (ROC indicator = 0.84).
- The uPA-T model generated a visualization tool for predicting cancer recurrence probability.
Conclusions:
- Non-linear KDA is a powerful approach for analyzing prognostic factors in breast cancer.
- The uPA-T model shows potential for clinical application in predicting recurrence.
- This method aids in understanding and visualizing patient-specific recurrence risks.