Breast Tumor Detection Using Robust and Efficient Machine Learning and Convolutional Neural Network Approaches.

Area of Science:

• Oncology
• Biomedical Engineering
• Data Science

Background:

• Breast cancer is the second most common cancer globally and a significant health concern, particularly for women over 50.
• While less common, breast cancer affects men and transgender individuals, highlighting the need for broad detection methods.
• Early detection of breast tumors is crucial for reducing mortality rates and improving patient outcomes.

Purpose of the Study:

• To develop and evaluate machine learning (ML) and deep learning (DL) models for accurate breast tumor identification.
• To identify critical prognostic indicators associated with breast cancer using predictive models.
• To assess the efficacy of various ML algorithms and a custom CNN for breast tumor detection.

Main Methods:

• Utilized a public dataset of breast tumor features for model development.
• Implemented and compared six models: Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Voting Classifier (VC), Support Vector Machine (SVM), and a custom Convolutional Neural Network (CNN).
• Evaluated model performance based on accuracy and F1 scores.

Main Results:

• The custom CNN model demonstrated superior performance with an average accuracy of 99%.
• All evaluated models (LR, DT, RF, VC, SVM, CNN) achieved high accuracy rates ranging from 96% to 99%.
• The F1 scores indicated strong predictive power, with the CNN achieving 99% and RF achieving 98%.

Conclusions:

• Machine learning and deep learning techniques show significant potential as accurate prognostic tools for breast tumor detection.
• The high accuracy of the developed models, especially the custom CNN, supports their use in clinical settings.
• These findings suggest that ML-based approaches can serve as valuable alternative tools for breast tumor detection, particularly in regions like Asia.