Data mining techniques in breast cancer diagnosis at the cellular-molecular level

Area of Science:

• Medical Informatics
• Computational Biology
• Oncology

Background:

• Machine learning and data mining offer promising avenues for improving breast cancer diagnosis.
• Early detection of breast cancer characteristics is crucial for effective treatment strategies.
• Reducing human error in medical diagnosis through accurate and rapid screening is a key objective.

Purpose of the Study:

• To develop a data mining-based method for early breast cancer detection.
• To minimize human errors in diagnosis through accurate and rapid screening.
• To aid in the better treatment of breast cancer by identifying characteristics in early stages.

Main Methods:

• The proposed method involves data pre-processing and image quality enhancement.
• Image segmentation is used to separate cancer cells from healthy tissue and remove outliers.
• An ensemble classification model combining deep neural networks and majority voting is configured.

Main Results:

• The method was evaluated on two histopathological datasets of invasive ductal carcinoma.
• High-level features were extracted, achieving average accuracies of 92.65% and 93.34%.
• The system demonstrated high performance in rapid diagnosis and classification of breast cancer.

Conclusions:

• Combining deep neural networks and feature extraction enhances diagnostic accuracy for breast cancer.
• This approach aids specialists in diagnosis and increases patient survival chances.
• The study contributes to advancing early detection and treatment of invasive ductal carcinoma.