Related Experiment Video
Updated: Apr 6, 2026
07:41
Performing Data Mining And Integrative Analysis Of Biomarker in Breast Cancer Using Multiple Publicly Accessible Databases
Published on: May 17, 2019
9.7K
A distributed fusion framework for breast cancer recurrence prediction using MapReduce.
Prachi Shahare1,2, Abha Mahalwar2, Aniket K Shahade3
1Department of Computer Science and Engineering, Yeshwantrao Chavan College of Engineering (YCCE), Nagpur, Maharashtra, India.
Scientific Reports
|April 4, 2026
Summary
This study introduces a hybrid AI framework for predicting breast cancer recurrence, combining multiple advanced machine learning models. The framework enhances prediction accuracy and scalability for diverse, large-scale datasets, improving clinical decision-making.
Area of Science:
- Oncology
- Artificial Intelligence
- Data Science
Background:
- Breast cancer recurrence poses a significant clinical challenge, impacting survival and treatment strategies.
- Accurate early prediction is difficult due to data heterogeneity, imbalance, and distributed storage across institutions.
Purpose of the Study:
- To develop a scalable, hybrid AI framework for accurate breast cancer recurrence prediction.
- To address challenges of heterogeneous and distributed datasets in clinical settings.
Main Methods:
- A MapReduce-aligned hybrid framework integrating Spark-based Gradient Boosted Trees, denoising autoencoder (AE) latent representations, calibrated XGBoost, and deep tabular models (FT-Transformer, TabTransformer).
- Evaluation on SEER breast cancer recurrence and Wisconsin Diagnostic Breast Cancer datasets.
Main Results:
- The AE-augmented fusion framework and calibrated XGBoost achieved superior discrimination on the Wisconsin dataset (ROC-AUC 0.9954, MCC ≥ 0.981).
- On the SEER dataset, the fusion framework improved recall for sparse recurrence signals, while calibrated XGBoost balanced precision and stability.
- Fusion learning enhanced sensitivity and stability; calibrated XGBoost showed strong discrimination.
Conclusions:
- The proposed framework offers a scalable and reliable solution for individualized breast cancer recurrence risk prediction.
- Combining diverse AI techniques (tree-based, AE, transformers) improves prediction performance and robustness.
- Spark-GBT integration ensures suitability for multi-institutional data without centralization.
