Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bridging Accuracy and Interpretability: Deep Learning with XAI for Breast Cancer Detection

Bishal Chhetri, B. V. Rathish Kumar

TL;DR

This work tackles the dual challenge of high-accuracy breast cancer detection and interpretability by applying an interpretable deep neural network to quantitative FNA-derived features from the UW breast cancer dataset. The authors compare a range of classical ML models and an ANN, with the 2-hidden-layer ReLU ANN trained via Adam achieving a peak accuracy of $0.992$ and precision of $1.000$, while maintaining strong recall ($0.977$) and F1 ($0.988$). To address the black-box concern, SHAP and LIME explain predictions at both global and local levels, revealing concave_points_mean as the most influential feature for malignancy. The results demonstrate a compelling combination of predictive performance and transparency, supporting potential clinical deployment and underscoring directions for future work to integrate additional clinical and genetic data for even more robust, explainable breast cancer diagnostics.

Abstract

In this study, we present an interpretable deep learning framework for the early detection of breast cancer using quantitative features extracted from digitized fine needle aspirate (FNA) images of breast masses. Our deep neural network, using ReLU activations, the Adam optimizer, and a binary cross-entropy loss, delivers state-of-the-art classification performance, achieving an accuracy of 0.992, precision of 1.000, recall of 0.977, and an F1 score of 0.988. These results substantially exceed the benchmarks reported in the literature. We evaluated the model under identical protocols against a suite of well-established algorithms (logistic regression, decision trees, random forests, stochastic gradient descent, K-nearest neighbors, and XGBoost) and found the deep model consistently superior on the same metrics. Recognizing that high predictive accuracy alone is insufficient for clinical adoption due to the black-box nature of deep learning models, we incorporated model-agnostic Explainable AI techniques such as SHAP and LIME to produce feature-level attributions and human-readable visualizations. These explanations quantify the contribution of each feature to individual predictions, support error analysis, and increase clinician trust, thus bridging the gap between performance and interpretability for real-world clinical use. The concave points feature of the cell nuclei is found to be the most influential feature positively impacting the classification task. This insight can be very helpful in improving the diagnosis and treatment of breast cancer by highlighting the key characteristics of breast tumor.

Bridging Accuracy and Interpretability: Deep Learning with XAI for Breast Cancer Detection

TL;DR

This work tackles the dual challenge of high-accuracy breast cancer detection and interpretability by applying an interpretable deep neural network to quantitative FNA-derived features from the UW breast cancer dataset. The authors compare a range of classical ML models and an ANN, with the 2-hidden-layer ReLU ANN trained via Adam achieving a peak accuracy of and precision of , while maintaining strong recall () and F1 (). To address the black-box concern, SHAP and LIME explain predictions at both global and local levels, revealing concave_points_mean as the most influential feature for malignancy. The results demonstrate a compelling combination of predictive performance and transparency, supporting potential clinical deployment and underscoring directions for future work to integrate additional clinical and genetic data for even more robust, explainable breast cancer diagnostics.

Abstract

In this study, we present an interpretable deep learning framework for the early detection of breast cancer using quantitative features extracted from digitized fine needle aspirate (FNA) images of breast masses. Our deep neural network, using ReLU activations, the Adam optimizer, and a binary cross-entropy loss, delivers state-of-the-art classification performance, achieving an accuracy of 0.992, precision of 1.000, recall of 0.977, and an F1 score of 0.988. These results substantially exceed the benchmarks reported in the literature. We evaluated the model under identical protocols against a suite of well-established algorithms (logistic regression, decision trees, random forests, stochastic gradient descent, K-nearest neighbors, and XGBoost) and found the deep model consistently superior on the same metrics. Recognizing that high predictive accuracy alone is insufficient for clinical adoption due to the black-box nature of deep learning models, we incorporated model-agnostic Explainable AI techniques such as SHAP and LIME to produce feature-level attributions and human-readable visualizations. These explanations quantify the contribution of each feature to individual predictions, support error analysis, and increase clinician trust, thus bridging the gap between performance and interpretability for real-world clinical use. The concave points feature of the cell nuclei is found to be the most influential feature positively impacting the classification task. This insight can be very helpful in improving the diagnosis and treatment of breast cancer by highlighting the key characteristics of breast tumor.
Paper Structure (15 sections, 7 equations, 15 figures, 4 tables)

This paper contains 15 sections, 7 equations, 15 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Data and Data Pre-processing
  3. Dataset Used
  4. Data Scaling and Feature Selection
  5. Methodology Followed
  6. Results
  7. Logistic Regression Prediction
  8. Decision Tree Prediction
  9. Random forest Prediction
  10. KNN Prediction
  11. XGBoost
  12. SGD Algorithm
  13. ANN Results
  14. Explanation of Model Prediction using SHAP and LIME
  15. Discussion and Conclusion

Figures (15)

  • Figure 1: Number of benign and malignant cases in the dataset.
  • Figure 2: Feature extraction workflow from a digitized cancer image.
  • Figure 3: First few columns of the original dataset.
  • Figure 4: Flowchart of the proposed methodology. The raw numerical data obtained from the feature extraction process were pre-processed and subsequently divided into training and testing sets prior to model training.
  • Figure 5: Feature importance in decision tree classification. The decision tree classifier identifies perimeter$\_$worst as the most important feature for classifying cancer as benign or malignant.
  • ...and 10 more figures