BCCHI-HCNN: Breast Cancer Classification from Histopathological Images Using Hybrid Deep CNN Models

Area of Science:

• Oncology
• Computer Science
• Medical Imaging

Background:

• Breast cancer is a leading global health concern, with 2.3 million new cases annually.
• Accurate and early diagnosis is crucial for effective treatment and improved patient outcomes.
• Histopathological image analysis presents opportunities for advanced diagnostic tools.

Purpose of the Study:

• To develop and evaluate a deep convolutional neural network (CNN) model for precise early breast cancer detection.
• To enhance the CNN model's performance by integrating multiple classifiers, including support vector machine (SVM), decision tree, and K-nearest neighbors (KNN).
• To assess the impact of feature vector selection, with and without principal component analysis (PCA), on diagnostic accuracy.

Main Methods:

• Utilized a deep convolutional neural network (CNN) architecture for image analysis.
• Integrated transfer learning techniques with SVM, decision tree, and KNN classifiers.
• Evaluated two feature vector approaches: with and without principal component analysis (PCA).
• Compared model performance against existing deep learning models using metrics like accuracy, precision, recall, and false/true positive rates.

Main Results:

• The SVM algorithm combined with PCA features demonstrated high accuracy (99.5%) and speed.
• The decision tree model achieved a high accuracy of 99.4% without PCA, though it was slower than SVM.
• Both models showed significant potential in accurately classifying breast cancer from histopathological images.

Conclusions:

• The proposed CNN-based model, particularly with SVM and PCA, offers a viable strategy for enhancing early breast cancer diagnosis.
• This approach can lead to more effective treatment decisions and improved patient survival rates.
• The study highlights the potential of integrating advanced machine learning techniques in medical diagnostics.