Radiomics-integrated machine learning framework for quantitative breast cancer diagnosis

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer is a leading cause of death in women globally.
• Early detection significantly improves treatment outcomes.
• Automated diagnostic tools are needed to enhance efficiency and accuracy.

Purpose of the Study:

• To develop an automated pipeline for breast tumor segmentation and classification using deep learning and radiomics.
• To improve the accuracy and speed of breast cancer diagnosis from ultrasound images.
• To reduce human intervention in the diagnostic process.

Main Methods:

• Utilized the UNet model for automated tumor segmentation in ultrasound scans.
• Extracted radiomic features from segmented regions for quantitative analysis.
• Employed machine learning algorithms for benign versus malignant tumor classification.

Main Results:

• The UNet-Radiomics-ML framework achieved a mean Intersection over Union (IoU) of 0.94231 for segmentation.
• The classification model demonstrated a high accuracy of 97.8% on testing data.
• Performance surpassed benchmark results from initial phase analysis.

Conclusions:

• The automated pipeline shows significant potential to enhance breast cancer diagnosis.
• Integration of deep learning, radiomics, and machine learning streamlines the diagnostic workflow.
• The proposed method offers a reliable and efficient approach for identifying malignant tumors.