Deep learning for differentiating benign from malignant tumors on breast-specific gamma image

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Oncology
• Computational Pathology

Background:

• Breast diseases pose a significant health risk to women globally.
• Accurate diagnosis of benign and malignant breast tumors is crucial for effective treatment.
• Increasing volumes of breast-specific gamma imaging (BSGI) data necessitate advanced diagnostic tools.

Purpose of the Study:

• To develop and evaluate a deep learning model for diagnosing breast tumors.
• Utilize breast-specific gamma imaging (BSGI) data as input for tumor classification.
• Compare the performance of the proposed deep learning model against other neural network architectures.

Main Methods:

• A dataset of 144 benign and 87 malignant tumors was curated and split into training (80%) and testing (20%) sets.
• A convolutional neural network, ResNet18, was implemented to create a novel deep learning diagnostic model.
• Model performance was assessed using accuracy, specificity, sensitivity, and Receiver Operating Characteristic (ROC) curves.

Main Results:

• The proposed deep learning model achieved superior performance with 99.1% accuracy, 98.8% specificity, and 99.3% sensitivity.
• The model outperformed traditional neural network and autoencoder models in diagnostic accuracy.
• Grad-CAM visualization was employed to enhance the interpretability of the deep learning model's diagnostic decisions.

Conclusions:

• The developed deep learning method offers a fast and reliable approach for physicians to diagnose breast tumors.
• This AI-driven tool has the potential to significantly improve diagnostic workflows in breast imaging.
• The study highlights the efficacy of deep learning in analyzing BSGI data for improved breast cancer detection.