Mammogram mastery: Breast cancer image classification using an ensemble of deep learning with explainable artificial...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer is a leading cause of cancer deaths in women globally.
• Early detection significantly improves survival rates.
• Manual mammogram analysis is time-consuming and prone to expert variability.

Purpose of the Study:

• To develop and evaluate an effective deep learning (DL) method for classifying mammogram images.
• To improve the accuracy and efficiency of breast cancer diagnosis using automated systems.
• To enhance the transparency of the classification process using interpretable AI.

Main Methods:

• Utilized a deep learning model pretrained on the Inception V3 architecture.
• Performed 5-fold cross-validation on a fully trained and fine-tuned Inception V3 model.
• Applied a combined method based on likelihood and mean for classification, incorporating interpretable AI techniques.

Main Results:

• The fine-tuned Inception V3 model achieved 99% accuracy and 99% F1 score in classifying mammogram images.
• The DL model demonstrated superior performance in distinguishing between cancerous and non-cancerous images.
• Interpretable AI techniques confirmed the model's effectiveness and transparency.

Conclusions:

• The proposed DL-based method is highly effective for automated breast cancer detection from mammogram images.
• AI-based solutions show significant potential to increase the accuracy and reliability of breast cancer diagnosis.
• This approach can enhance image-based diagnostic methods, supporting clinical decision-making.