Scholarlink

Purpose <p>This study investigates breast cancer diagnosis using radiomic features, Genetic Algorithms (GA) for feature selection, and Explainable AI (XAI) techniques.</p> Method <p>Local Interpretable Model-Agnostic Explanations (LIME) were applied to enhance the interpretability of machine learning (ML) models e.g., Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), and XGBoost. GA was used to select optimal radiomic features, improving predictive accuracy and computational efficiency.</p> Results <p>GA significantly improved classifier performance. RF precision increased from 82.69% to 85.26%, and NB from 66.67% to 81.41%. Accuracy improvements were also observed in LR, SVM, and XGBoost. GA-enhanced models achieved higher LIME fidelity scores, with NB improving from 66.42 to 99.89, indicating more reliable interpretability.</p> Conclusion <p>Integrating radiomic features with GA-optimized classifiers enhances breast cancer prediction. LIME provides essential model transparency, supporting clinical adoption by clarifying decision-making processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrating genetic algorithms and explainable AI for radiomics-based breast cancer detection

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Zulfkar Ali Ansari,
  • Hemlata Pant,
  • Ahmed Khan,
  • Susmitha Uddaraju,
  • P. Venkata Hari Prasad,
  • Aaliya Sarfaraz

摘要

Purpose

This study investigates breast cancer diagnosis using radiomic features, Genetic Algorithms (GA) for feature selection, and Explainable AI (XAI) techniques.

Method

Local Interpretable Model-Agnostic Explanations (LIME) were applied to enhance the interpretability of machine learning (ML) models e.g., Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), and XGBoost. GA was used to select optimal radiomic features, improving predictive accuracy and computational efficiency.

Results

GA significantly improved classifier performance. RF precision increased from 82.69% to 85.26%, and NB from 66.67% to 81.41%. Accuracy improvements were also observed in LR, SVM, and XGBoost. GA-enhanced models achieved higher LIME fidelity scores, with NB improving from 66.42 to 99.89, indicating more reliable interpretability.

Conclusion

Integrating radiomic features with GA-optimized classifiers enhances breast cancer prediction. LIME provides essential model transparency, supporting clinical adoption by clarifying decision-making processes.