Categorized contrast enhanced mammography dataset for diagnostic and artificial intelligence research

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Contrast-enhanced spectral mammography (CESM) offers higher diagnostic accuracy than digital mammography (DM).
• Deep learning (DL) models show promise for breast cancer detection, but existing models are trained on DM images due to a lack of CESM datasets.
• A significant gap exists in AI development for CESM analysis.

Purpose of the Study:

• To introduce the first comprehensive dataset for CESM images, named Categorized Digital Database for Low energy and Subtracted Contrast Enhanced Spectral Mammography images (CDD-CESM).
• To facilitate the development and evaluation of AI-driven decision support systems for CESM analysis.
• To address the limitation of insufficient training data for DL models in CESM.

Main Methods:

• Compilation of 2006 CESM images with detailed annotations, including segmentation masks, medical reports, and pathological diagnoses.
• Dataset categorization into various findings: mass, architectural distortion, asymmetry, calcifications, mass enhancement, non-mass enhancement, postoperative, and normal images.
• Development and evaluation of a DL-based technique for automatic segmentation of abnormal findings within CESM images.

Main Results:

• The CDD-CESM dataset comprises 2006 images with diverse pathological findings and comprehensive annotations.
• The dataset includes 248 images with multiple findings, offering a complex resource for AI model training.
• The proposed DL technique demonstrated effectiveness in automatically segmenting abnormal findings in CESM images.

Conclusions:

• The release of the CDD-CESM dataset is a significant contribution to advancing AI in breast cancer diagnostics using CESM.
• This dataset will accelerate the development of robust DL models for improved accuracy in CESM interpretation.
• The availability of annotated CESM data is crucial for enhancing AI-based decision support systems in radiology.